Parallel, adaptive delta sigma ADC

[0001] This application is a Non-Provisional application including the subject matter and claiming the priority date under 35 U.S.C. §119(e) of Provisional Application Ser. No. 60/607,577, filed Sep. 8, 2004, the contents of which are meant to be incorporated herein by reference.

[0002] 1. Field of the Invention

[0003] The present invention relates generally to analog-to-digital converters and control structures and methods for an analog-to-digital converter.

[0004] 2. Description of the Related Art

[0005] Analog-to-digital (ADC) converters are essential components in today's electronic circuits and systems. ADC converters transform analog signals to digital signals. Conventional delta sigma (ΔΣ) analog-to-digital converters offer high resolution and linearity, high integration, little differential non-linearity, and low cost. Their performance is not limited by mismatched components within the converter, and has low noise sensitivity. Most of the circuitry in delta sigma ADCs is digital; hence the performance of delta sigma ADCs does not drift with time and temperature.

[0006] Two basic principles govern the operation of conventional delta sigma ADCs: oversampling and noise shaping. The sampling frequency in a delta sigma ADC is typically chosen to be much larger than the input signal bandwidth. Oversampling spreads the quantization noise power over a bandwidth equal to the sampling frequency. A delta sigma ADC usually contains a delta sigma modulator, a lowpass filter, and a decimator filter. The delta sigma modulator applies a lowpass filter to the input analog signal and a high pass filter to the noise, hence placing most quantization noise energy above the input signal bandwidth. The lowpass filter follows the delta sigma modulator, attenuating out-of-band quantization noise. The decimator filter downsamples the sampled output digital signal to the Nyquist rate.

[0007] Since delta sigma ADCs typically operate at an oversampled rate much larger than the maximum input signal bandwidth, their circuitry is complex and their speed is low. Because of speed limitations, delta sigma ADCs perform best in high-resolution, very-low frequency applications.

[0008] In order to successfully extend the use of delta sigma ADCs to higher frequency applications, a parallel delta sigma ADC architecture has been proposed. U.S. Pat. No. 5,196,852 by Ian Galton and “A Nyquist-Rate Delta-Sigma A/D Converter” IEEE Journal of Solid-State Circuits, Vol. 33, No. 1, pp. 45–52 describe parallel delta sigma ADC systems. The system described in U.S. Pat. No. 5,196,852 achieves an effective oversampling ratio of N*M, where N is the oversampling ratio of each delta sigma ADC and M is the number of parallel delta sigma ADC channels. The system described in “A Nyquist-Rate Delta-Sigma A/D Converter” achieves an effective oversampling ratio of M without oversampling in the individual delta sigma ADCs, where M is the number of parallel delta sigma ADC channels. However, with the circuits described in above works, the parallel delta sigma ADCs do not self-adapt. Hence, only a limited predetermined range of incoming signal frequencies can be processed.

[0009] A disclosed embodiment of the application addresses these and other issues by utilizing a parallel, adaptive delta sigma analog-to-digital converter.

[0010] Embodiments of the present invention are directed to an apparatus for adaptive analog-to-digital conversion. According to a first aspect of the present invention, an apparatus for adaptive analog-to-digital conversion comprises: a frequency modulator unit for changing an input analog signal into a modulated analog signal with a frequency spectrum in a bandwidth of interest; a parallel delta sigma conversion unit operatively connected to the frequency modulator unit, the parallel delta sigma conversion unit converting the modulated analog signal into a digital signal; and a controller operatively connected to the frequency modulator unit and the parallel delta sigma conversion unit, the controller adjusting at least one parameter relating to a frequency characteristic of the frequency modulator unit and/or the parallel delta sigma conversion unit.

[0033] Aspects of the invention are more specifically set forth in the accompanying description with reference to the appended figures.

[0034] FIG. 1 is a block diagram of a parallel, adaptive delta sigma ADC according to an embodiment of the present invention. The parallel, adaptive delta sigma ADC 100 illustrated in FIG. 1 includes the following components: a frequency modulator unit 20; a delta sigma conversion unit 50; and a control unit 70 operatively connected as shown. An input analog signal 1 enters the parallel, adaptive delta sigma ADC 100 through frequency modulator unit 20. Parallel, adaptive delta sigma ADC 100 converts input analog signal 1 to output digital signal 90. Control unit 70 controls the operation of frequency modulator 20 and delta sigma conversion unit 50. Delta sigma conversion unit 50 may also communicate with control unit 70 before outputting digital signal 90. The parallel, adaptive delta sigma ADC 100 may be built for automatic, manual, or selectable automatic/manual control of output digital signal 90. Operation of the parallel, adaptive delta sigma ADC 100 in FIG. 1 will become apparent from the following discussion.

[0035] FIG. 2A illustrates a frequency modulator unit 201 that may be included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Frequency modulator unit 201 contains a tunable bandpass filter 140 operatively connected as shown. Frequency modulator unit 201 may also contain a local oscillator (LO) 110 and a mixer 120 operatively connected as shown. The LO 110 and mixer 120 are needed when input analog signal 1 lies outside the bandwidth accessible to delta sigma conversion unit 50. For some embodiments, frequency modulator unit 201 may contain a chain of local oscillators 110. Control unit 70 controls tunable bandpass filter 140. Control unit 70 may also control local oscillator 110.

[0036] Local oscillator 110 is a semiconductor device or an electronic circuit that generates one or more signals of constant frequency. The frequencies generated by local oscillator 110 are determined by parameters of electronic components inside local oscillator 110 as is known in the art. Mixer 120 is a semiconductor device or electronic circuit that multiplies two signals of different frequencies to obtain a signal of intermediate frequency. A frequency υ_LO generated by local oscillator 110 is mixed by mixer 120 with all frequencies contained in input analog signal 1, producing mixed analog signal 3. Each frequency υ_in present in input analog signal 1 is shifted to two frequencies υ_LO−υ_in and υ_LO+υ_in mixed analog signal 3.

[0037] Local oscillator 110 is preferably tunable. This may be accomplished with, for example, a single tunable local oscillator 110 or a chain of tunable local oscillators 110 as is known in the art.

[0038] Tunable bandpass filter 140 is a semiconductor device or an electronic circuit that isolates and extracts independent communication channels from the frequency wideband of mixed analog signal 3, and cuts off frequencies in the frequency wideband of mixed analog signal 3 that are either too high or too low. Tunable bandpass filter 140 may be a transmultiplexer; a spectral subband coder that divides an analog signal into frequency segments, or spectral terms, computed for non-overlapped successive blocks of input data; a collection of single-sideband narrowband filters which perform a complex heterodyne, hence basebanding a selected center frequency; a channelizer with suitable RF switches; an antialias filter; a band pass filter; a lowpass filter; or a bandpass and lowpass filter. The center frequency and/or bandwidth of tunable bandpass filter 140 can be changed manually or automatically, so that the passband of tunable bandpass filter 140 matches the bandwidth of interest of delta sigma conversion unit 50, hence eliminating aliasing.

[0039] FIG. 2B illustrates aspects of the operation of a local oscillator 110 and a mixer 120 that may be included in a frequency modulator unit 201 according to an embodiment of the present invention. In case input analog signal 1 occupies a different frequency band than the frequency band parallel, adaptive delta sigma ADC 100 is designed for, the local oscillator 110 may be tuned such that the input signal is downshifted onto a desired bandwidth range.

[0040] In other words, the particular design values and construction of the parallel, adaptive delta sigma ADC 100 result in a device having a particular bandwidth range (bandwidth of interest). The local oscillator 10 may be used to shift the input signal onto the bandwidth range of the ADC 100. For this purpose, local oscillator 10 may be tuned by control unit 70 to generate a signal of frequency υ_LO which is mixed with all frequencies of input analog signal 1 in mixer 120, producing mixed analog signal 3. Each frequency υ_in present in input analog signal 1 is shifted to two frequencies υ_LO−υ_in and υ_LO+υ_in in mixed analog signal 3 as illustrated in FIG. 2B. υ_LO is chosen so that either υ_LO−υ_in or υ_LO+υ_in is located within the frequency band for which the parallel, adaptive delta sigma ADC 100 is designed.

[0041] FIG. 2C illustrates aspects of the operation of a tunable bandpass filter 140 included in a frequency modulator unit 201 according to an embodiment of the present invention. Tunable bandpass filter 140 divides the frequency wideband of mixed analog signal 3 into frequency segments 5A, 5B, . . . 5Y, 5Z. The number of frequency segments and the width of frequency segments may be tuned by control unit 70. All frequency segments are then mapped to the lowest frequency segment 5Z, avoiding aliasing of frequency segments.

[0042] FIG. 3 illustrates a delta sigma conversion unit 501 that may be included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Delta sigma conversion unit 501 includes a code generator 150, N number of delta sigma (ΔΣ) channels 155₁, 155₂, . . . 155_N connected in parallel, and an adder 260 operatively connected as shown. Each ΔΣ channel 155_i includes an input multiplier 160_i, a lowpass ΔΣ ADC 180_i, an adaptable digital correction filter 200_i, a programmable decimation filter 220_i, and an output multiplier 240_i, where subscript “i” has values from 1 to N.

[0043] A lowpass ΔΣ ADC 180_i is a delta sigma analog-to-digital converter and may be implemented using many possible converter types including low-order single-bit single-loop converter, high-order single-bit single-loop converter, single-bit multiloop (cascaded or MASH) converter and multi-bit (single-loop or multiloop) converter. All lowpass ΔΣ ADCs 180_i (for i from 1 to N) are preferably substantially identical. In one embodiment of the current invention, all lowpass ΔΣ ADC 180_i have the same order, number of bits, and signal delay.

[0044] Adaptable digital correction filters 200_i, programmable decimation filters 220_i, and output multipliers 240_i where subscript “i” has values from 1 to N, are conventional digital electronic devices. They can be implemented using a custom ASIC, an off the shelf FIR filter, a field programmable gate array, or a sufficiently fast microprocessor. The filter coefficients for the adaptable digital correction filters 200_i and the programmable decimation filters 220_i, may be stored in a register on the chip that holds the parallel, adaptive delta sigma ADC 100, or in a separate memory off the chip, connected to a data bus. Adaptable digital correction filters 200_i, programmable decimation filters 220_i, and output multipliers 240_i, where subscript “i” has values from 1 to N, may also be combined into a single filter.

[0045] As will be further described below, adaptable digital correction filters 200_i and programmable decimation filters 220_i have programmable filter parameters such as length and bandwidth. Adaptable digital correction filters 200_i correct inaccuracies introduced by lowpass ΔΣ ADCs 180_i where subscript “i” has values from 1 to N, and may be implemented using known techniques.

[0046] Programmable decimation filters 220_i are used to down-sample signals and to eliminate out-of-band noise, and may be implemented using known techniques. Code generator 150 generates a set of codes. Input multipliers 160_i and output multipliers 240_i are standard multiplier circuits whose multiplication values are supplied by codes from the set of codes provided by code generator 150. Adder 260 may be a simple adder circuit that adds its inputs bit-by-bit. Control unit 70 may control code generator 150, adaptable digital correction filters 200_i, and programmable decimation filters 220_i.

[0047] A power splitter 159 simultaneously inputs one of the frequency segments produced by tunable bandpass filter 140 (signal 5A) to all ΔΣ channels 155₁, 155₂, . . . 155_N. The set of codes generated by code generator 150 and applied by multipliers 160_i decomposes signal 5A into orthonormal components. Along with other types of codes, a set of Hadamard codes or a set of Gold codes can be generated by code generator 150 for orthonormal decomposition of signal 5A. Each input multiplier 160_i uses one and only one code from the set of codes generated by code generator 150, to create a multiplication value for its frequency decomposing weighting function. Frequency decomposing weighting functions are orthonormal and channel specific. Inside channel i, signal 5A is frequency-decomposed by input multiplier 160_i, passed through lowpass ΔΣ ADC 180_i, corrected by adaptable digital correction filter 200_i for inaccuracies gained from lowpass ΔΣ ADC 180_i, down-sampled by programmable decimation filter 220_i which also eliminates quantization noise, and multiplied by output multiplier 240_i which applies a time-shifted version of the frequency decomposing weighting function of input multiplier 160_i, that undoes the frequency decomposing action of input multiplier 160_i.

[0048] The outputs from all ΔΣ channels 155 are summed by adder 260 to obtain a digital signal 7.

[0049] If signal 5A has a frequency spectrum of X GHz, each lowpass ΔΣ ADC 180_i may be clocked at 2× GHz, which is the Nyquist rate of signal 5A frequency spectrum. Each lowpass ΔΣ ADC 180_i is designed to have a band-pass response of X/N, hence the widest bandwidth signal the N-channel parallel, adaptive delta sigma ADC 100 can digitize is X GHz (N*X/N). The oversampling rate (OSR) of delta sigma conversion unit 501 for the entire frequency band of signal 5A is 1, since each lowpass ΔΣ ADC 180_i is clocked at the Nyquist rate of signal 5A frequency spectrum. However, within each ΔΣ channel 155_i the OSR ratio is N, calculated as the ratio of each lowpass ΔΣ ADC 180_i sampling frequency (2×) to the Nyquist frequency of the band-pass response of the channel (2*X/N). Therefore the parallel network of individual ΔΣ channels 155_i has N times larger OSR than a single lowpass ΔΣ ADC 180_i would exhibit.

[0050] FIG. 4A illustrates a code generator 150A that may be included in a delta sigma conversion unit 501 of a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Code generator 150A is a code calculator that generates a set of codes in real-time, without any information except the desired length of the set of codes which is supplied by control unit 70. Each code C_i in the set of codes is sent to a pair of input multipliers 160_i and output multipliers 240_i. Input multipliers 160_i and output multipliers 240_i generate multiplication values from the code C_i sent to them.

[0051] To account for processing times of the ΔΣ ADC 180_i, adaptable digital correction filter 200_i, and programmable decimation filter 220_i, the codes supplied to output multipliers 240_i are preferably time shifted or delayed relative to the input multipliers 160_i. This may be accomplished by, for example, a delay element (not shown), by the code calculator generator 150A, or by the control unit 70.

[0052] FIG. 4B illustrates a code generator 150B that may be included in a delta sigma conversion unit 501 of a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Code generator 150B includes a memory 146 with registers or memory locations 148₁, 148₂, . . . 148_N that contain a set of codes. Code generator 150B generates a set of output codes C₁, C₂ . . . C_N by direct read from memory registers or memory locations 148₁, 148₂, . . . 148_N. The code C_i in each register 148_i is directly read into input multiplier 160_i and output multiplier 240_i. Input multipliers 160_i and output multipliers 240_i generate multiplication values from code C_i.

[0053] FIG. 4C illustrates a code generator 150C that may be included in a delta sigma conversion unit 501 of a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Code generator 150C includes a memory 147 with M sets of registers or memory locations. Each register set “j”, with “j” from 1 to M, contains N individual registers or memory locations 148_1j, 148_2j, . . . 148_Nj. Each register 148_ij stores an internal code H_ij. There are N×M internal codes H_ij as “i” runs from 1 to N and “j” runs from 1 to M. A switchable lookup table register 143 decides output codes C₁, C₂ . . . C_N by direct read from among all internal codes H_ij. Output codes C₁, C₂ . . . . C_N are sent to all N ΔΣ channels 155₁, 155₂, . . . 155_N to input multiplier 160_i and output multiplier 240_i where subscript “i” has values from 1 to N. Input multipliers 160_i and output multipliers 240_i generate multiplication values from the code sent to them. Lookup table register 143 may store one lookup table or multiple lookup tables. An example of a lookup table is:

[0054] $\left[\begin{array}{ccccc}{H}_{11}& H_{21}& \cdots & \cdots & H_{\mathrm{N1}}\\ \cdots & \cdots & \cdots & \cdots & \cdots \\ \cdots & \cdots & \cdots & \cdots & \cdots \\ H_{1M}& H_{2M}& \cdots & \cdots & H_{\mathrm{NM}}\end{array}\right].$
This table would be read by direct memory read, switching between rows to select one row for output codes C₁, C₂ . . . C_N.
An example of a multiple lookup table is:

[0055] $\hspace{1em}\left[\begin{array}{cccccccccccccc}{H}_{11}& H_{11}& \dots & H_{11}& H_{21}& H_{21}& \dots & H_{21}& \dots & \dots & H_{\mathrm{N1}}& H_{\mathrm{N1}}& \dots & H_{\mathrm{N1}}\\ H_{12}& H_{12}& \dots & H_{12}& H_{22}& H_{22}& \dots & H_{22}& \dots & \dots & H_{\mathrm{N2}}& H_{\mathrm{N2}}& \dots & H_{\mathrm{N2}}\\ \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots & \cdots \\ H_{1M}& H_{1M}& \dots & H_{1M}& H_{2M}& H_{2M}& \dots & H_{2M}& \dots & \dots & H_{\mathrm{NM}}& H_{\mathrm{NM}}& \dots & H_{\mathrm{NM}}\end{array}\right]$

[0056] FIG. 5 illustrates a control unit 700 that may be included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Control unit 700 may include the following operational components: a mode control unit 305; a bandwidth control unit 300; a self calibration unit 320; and a DAC 340. Bandwidth control unit 300 controls code generator 150, programmable decimation filters 220₁, . . . 220_N, tunable bandpass filter 140, and tunable local oscillator 110. Self calibration unit 320 controls adaptable digital correction filters 200₁, . . . 200_N and DAC 340, and communicates with bandwidth control unit 300.

[0057] The parallel, adaptive delta sigma ADC 100 has two modes of operation: a normal operation mode and a self-calibration mode. Mode control unit 305 determines the operation mode of the parallel, adaptive delta sigma ADC 100.

[0058] The parallel, adaptive delta sigma ADC 100 operates in normal operation mode most of the time. In normal operation mode, bandwidth control unit 300 determines the bandwidth of interest using internal algorithms or an external control signal manually or automatically commanded. Bandwidth control unit 300 calculates or looks up the proper code for code generator 150, and calculates filter coefficients for programmable decimation filters 220_i, where “i” runs from 1 to N. Bandwidth control unit 300 also sets the center frequency and the bandwidth for tunable bandpass filter 140 and tunable local oscillator 110. After sending commands to code generator 150, programmable decimation filters 220_i, tunable bandpass filter 140, and tunable local oscillator 110, bandwidth control unit 300 remains idle until mode control unit 305 sends a command to change bandwidths again.

[0059] In self-calibration mode, self calibration unit 320 takes over bandwidth control unit 300. Self calibration unit 320 sets filter coefficients for adaptable digital correction filters 200_i to zero, and then instructs bandwidth control unit 300 to set a predetermined calibration bandwidth. Self calibration unit 320 then creates a calibration signal using the built-in DAC 340. The DAC 340 feeds the calibration signal to tunable bandpass filter 140 as an input. The calibration signal is processed by tunable bandpass filter 140 in the same manner as in normal mode operation. The digital output of the delta sigma conversion unit 501 is fed back to self calibration unit 320. Self calibration unit 320 calculates a final set of adaptable digital correction filter 200_i coefficients for use when control unit 700 and parallel, adaptive delta sigma ADC 100 return to normal mode operation.

[0060] Mode control unit 305 may be a part of control unit 700. In another embodiment of the invention, mode control unit 305 may be external to the parallel, adaptive delta sigma ADC 100. The DAC 340 may be any conventional DAC. The DAC 340 function may also be generated by an alternate analog frequency synthesis technique. Mode control unit 305, bandwidth control unit 300, and self calibration unit 320 may be implemented using a single field programmable gate array; an ASIC; a microcontroller; a standard microprocessor; or a memory or set of memories.

[0061] FIG. 6 illustrates aspects of the operation of a control unit 700 that may be included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. Control unit 700 selects (902) the operation mode of the parallel, adaptive delta sigma ADC 100.

[0062] If the selected operation mode is the normal operation mode (904), control unit 700 determines (905) the frequency or frequencies of interest for the parallel, adaptive delta sigma ADC 100. Next, control unit 700 controls tunable local oscillator 110 by selecting (906) the frequency of tunable local oscillator 110; controls the tunable bandpass filter 140 by selecting the number of desired frequency segments (907), and/or the bandwidth of frequency segments (908), and/or the filter bandwidth (909); determines the length of code (910) and the type of code (911) for code generator 150; and controls programmable decimation filters 220₁ . . . 220_N by selecting the filters' function (912), cutoff frequency (913), and length (914). In one embodiment of the invention, code calculator generator 150A directly calculates a code according to a selected code length (910). In other embodiments of the invention, code generators 150B and/or 150C choose an appropriate lookup table and/or lookup table read rate according to the selected code length (910) and the type of code (911).

[0063] If the selected operation mode is the self-calibration mode (903), control unit 700 determines a calibration signal (915) for an appropriate calibration bandwidth, and sends it to DAC 340. Control unit 700 also controls the adaptable digital correction filters 200₁, . . . 200_N by calculating (916) proper filter correction coefficients.

[0064] FIG. 7 is a flow diagram illustrating operations performed by a frequency modulator unit 201 according to an embodiment of the present invention. An analog signal 1 (481) is input into frequency modulator unit 201. A test is performed (482) to determine whether input analog signal 1 lies outside the bandwidth accessible by delta sigma conversion unit 501. If input analog signal 1 lies outside the bandwidth accessible by delta sigma conversion unit 501, input analog signal 1 frequency is shifted to a frequency accessible to delta sigma conversion unit 501, using (483) tunable local oscillator 110 whose frequency is controlled (906) by control unit 700. Input analog signal 1 is then filtered (484) by tunable bandpass filter 140 that is controlled (1100) by control unit 700. The output of frequency modulator unit 201 is analog input signal 5A (615).

[0065] FIG. 8 is a flow diagram illustrating operations performed by a delta sigma conversion unit 501 according to an embodiment of the present invention. Delta sigma conversion unit 501 is designed to convert a bandwidth or set of separate bandwidths present in input analog signal 1 into a digital signal. Analog signal 5A (615) output from frequency modulator unit 201 has a continuous low frequency band or an arbitrary set of bandwidths.

[0066] Power splitter 159 splits (620) analog signal 5A into N identical signal channels. Each signal channel is passed to a ΔΣ channel 155_i, with subscript “i” having values from 1 to N. In each ΔΣ channel 155_i, input multiplier 160_i multiplies (830) the signal channel by a code from a code set (820) created by code generator 150. Power splitter 159 and input multipliers 160_i break the band or bands of interest of analog signal 5A into many continuous low frequency bands, by decomposing input analog signal 5A into several subbands in which frequencies of interest are translated bit-by-bit to low frequency in each of the ΔΣ channels 155_i.

[0067] Control unit 700 sets (750a) the type and length of the code set that may be generated by real time calculation (824) in a code calculator generator 150A, by direct read (822) from a memory 150B containing a lookup table, or by direct read (826) from a memory with switchable output 150C containing a group of code sets located in a memory lookup table. The generated code set is sent (820) to input multipliers 160_i and output multipliers 240_i.

[0068] The signal channel in each ΔΣ channel 155_i is then converted (832) to a digital signal channel by lowpass ΔΣ ADC 180_i. Errors introduced by lowpass ΔΣ ADCs 180_i are compensated (834) by the adaptable digital correction filters 200_i whose filter correction coefficients are set (750b) by control unit 700. Quantization noise is removed (836) by programmable decimation filters 220_i whose decimation filter coefficients are set (750c) by control unit 700.

[0069] The digital signal channel in each ΔΣ channel 155_i is demodulated (838) by output multiplier 240_i and is output from the ΔΣ channel 155_i. The output digital signal channels from all parallel ΔΣ channels 155_i for “i” from 1 to N are added (840) together by adder 260. The output (842) of the adder is digital signal 7.

[0070] For analog input signal 5A to be accurately digitally reconstructed from the outputs of programmable decimation filters 220_i, the coefficients of programmable decimation filters 220_i are adapted (750c) by control unit 700 to the specific code (820) that was generated by code generator 150. Hence, changing the code (820) generated by code generator 150 and the coefficients of programmable decimation filters 220_i with “i” from 1 to N, changes the band or bands in analog input signal 5A that can be processed by delta sigma conversion unit 501. Therefore, by changing the code generated by code generator 150 and the coefficients for programmable decimation filters 220_i, the parallel, adaptive delta sigma ADC 100 can change bandwidth and dynamic range. One set of codes and programmable decimation filters coefficients may create a wideband low dynamic range modulator, while another set of codes and programmable decimation filters coefficients may create a narrow band, high dynamic range modulator. The parallel ΔΣ channels 155_i may also be split into separate channel groups to create a modulator with separate pass-bands. In this case, separate bands of different bandwidth as well as continuous bands from analog signal 5A can be processed by the parallel, adaptive delta sigma ADC 100.

[0071] FIG. 9A is a flow diagram illustrating aspects of an exemplary normal operation mode of a control unit 700 included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. In a normal operation mode, bandwidth control unit 300 included in control unit 700 sets (1200) the tunable local oscillator 110, the tunable bandpass filter 140, the code generator 150, and the programmable decimation filters 220_i to proper settings for a frequency band or bands of interest. The frequency band or bands of interest can be selected by bandwidth control unit 300, or by an external command from an external signal. When signals of interest and consequently bands of interest are unknown, such as in passive listening of signals or in electronic signal warfare, bandwidth control unit 300 selects the frequency band or bands of interest using a self adaptive algorithm. When the input signals are known, such as in a RADAR signal emission, an external signal may direct the bandwidth control unit 300 to a specific frequency band with a desired dynamic range.

[0072] An exemplary self adaptive algorithm is presented in FIG. 9A. In one logic flow, the bandwidth control unit 300 sets (1202) all N parallel ΔΣ channels 155_i of delta sigma conversion unit 501 in use to form a wideband low dynamic range modulator to search for signals. Once a signal is detected (1204), bandwidth control unit 300 switches (1206) to a narrow band, high dynamic range configuration, to listen to the identified signal. This logic flow works well for large complex signals. A second logic flow may search for both large and small signals using (1208) half of the N parallel ΔΣ channels 155_i to form a wideband low dynamic range modulator to search for large signals, and half (1210) of the N parallel ΔΣ channels 155_i to form a narrow band, high dynamic range modulator to search for small signals. By slowly sweeping the center frequency of the high dynamic range band, bandwidth control unit 300 can locate (1212) small persistent signals. Once one or more signals of interest have been identified (1214), the bandwidth control unit 300 can allocate one or more parallel ΔΣ channels 155_i of the delta sigma conversion unit 501 to listen (1216) to the signals of interest, while using the remaining parallel ΔΣ channels 155_i to search (1218) for additional signals in other bands.

[0073] FIG. 9B is a flow diagram illustrating aspects of an exemplary self-calibration operation mode of a control unit 700 included in a parallel, adaptive delta sigma ADC 100 according to an embodiment of the present invention. In self-calibration mode, the small errors introduced by delta sigma modulators 180_i are quantified. Error quantification is important because it allows for error correction and compensation by adaptable digital correction filters 200_i. One exemplary implementation of error quantification and correction is shown in FIG. 9B, where a known signal is input into the delta sigma conversion unit 501, and the output of the delta sigma conversion unit 501 is examined to determine introduced errors.

[0074] When control unit 700 decides that it is time for a self calibration cycle (1230) based on some external input or some internal routine, it activates (1232) the self calibration unit 320, which instructs the bandwidth control unit 300 to switch (1234) to a predetermined calibration bandwidth. The bandwidth control unit 300 sets (1236) the code generator 150, tunable bandpass filter 140, and decimation filters 220_i, to the proper settings for the desired bandwidth. The self calibration unit 320 then activates the DAC 340 to produce (1238) a known calibration signal which is injected (1242) into the delta sigma conversion unit 501. The coefficients on the digital correction filters 200_i are all set to zero (1240) so that no digital correction is performed. The uncorrected digital output signal 7 (1244) is fed back (1246) to the self-calibration unit and analyzed to determine what errors (1248) are introduced by the low pass ΔΣ ADCs 180_i. The self-calibration unit uses this signal to calculate (1250) a new set of coefficients (1254) for the adaptable digital correction filters 200_i with “i” from 1 to N.

[0075] Optionally, the calibration signal could be reinjected into delta sigma conversion unit 501 (1242) using the most recent set of adaptable digital correction filters 200_i coefficients. The output is re-measured to determine if errors are within tolerance (1249). Such errors are primarily nonlinearities and variations in gain from one ΔΣ channels 155_i to another. If the errors are within tolerance, the self calibration is completed (1252) and the latest found set of coefficients for adaptable digital correction filters 200_i with “i” from 1 to N, are stored to be used (1254) until the next self calibration cycle. Other variations on the self-calibration routine include calibrating each channel one at a time, and calibrating at various frequencies to eliminate frequency dependent errors.

[0076] FIG. 10 illustrates an example of code generator 150 (Hadamard code generator 1500) that may be included in a delta sigma conversion unit 501 according to an embodiment of the present invention. The orthogonal set of codes generated by a Hadamard code generator 1500 is used in a preferred embodiment of the current invention. A Hadamard code generator 1500 generates a Hadamard code from a conventional Hadamard matrix. The elements of a Hadamard matrix are 1s and −1s. The size of the Hadamard matrix used by Hadamard code generator 1500 is N×N, where N is the number of ΔΣ channels 155 in delta sigma conversion unit 50. In order for the Hadamard matrix to exist, N can be a non-negative power of 2, or another number for which Hadamard matrices are known to exist. Such numbers may include all multiples of 4 smaller than 428. For N=1 the Hadamard matrix is H₁=[1]. For N≧2, Hadamard matrix of order N, H_N, is defined recursively as follows:

[0077] $H_N=\left[\begin{array}{cc}{H}_{N/2}& H_{N/2}\\ H_{N/2}& -H_{N/2}\end{array}\right].$
The N-by-N Hadamard matrix has the property that H_NH_N^T=NI_N where I_N is the N-by-N identity matrix. The rows and columns of the Hadamard matrix are mutually orthogonal. The Hadamard codes 151 generated by Hadamard code generator 1500 are the individual rows of the N×N Hadamard matrix. The orthogonal set of codes 151 is sent to input multipliers 160 and to output multipliers 240.

[0078] FIG. 11 illustrates a continuous time integrator circuit 400 that may be included in a ΔΣ ADC 180_i from a delta sigma conversion unit 501, according to an embodiment of the present invention. An exemplary embodiment of this invention uses a continuous time 4^th order cascade of resonator filters that exhibits high stability. A building block of the filter is a continuous time integrator 400 illustrated in FIG. 8. A transconductor stage charging a capacitor 440 implements the integrating function. A metal film resistor 420 is used in place of a transconductance element to improve and maintain the high degree of linearity required for high dynamic range performance of ΔΣ ADCs 180. The schematic in FIG. 11 is a single ended design. A fully differential design such as that shown in FIG. 12 can also be used.

[0079] FIG. 12 illustrates a polarity reversal circuit 500 that may be included in input multipliers 160_i and output multipliers 240_i of a delta sigma conversion unit 501 according to an embodiment of the present invention. The polarity reversing circuit 500 reverses the differential input lines 520 and 521 to ΔΣ ADCs 180 at the appropriate times when the +/−1 Hadamard input multiplier sequences 152 and Hadamard output multiplier sequences 153 are applied to signal 5A. An exemplary implementation of polarity reversal circuit 500 uses current steering bipolar switches 540 in emitter-follower unity gain buffers connected at the input of continuous time integrator circuit 400 of the first resonator in the continuous time 4^th order cascade of resonators filter in ΔΣ ADCs 180. Bipolar transistors perform well as high-speed current switches and the negative feedback of the emitter-follower buffer configuration maintains circuit linearity. The input lines 560 and 561 labeled H+ and H− in the circuit diagram of FIG. 9 correspond to logical levels of 1 and −1 in a differential Hadamard input multiplier sequence 152 or Hadamard output multiplier sequence 153. Polarity reversal circuit 500 is used to switch the inputs to the integrator 400, between input value and inverted input value. Polarity reversal circuit 500 combines input multiplier 160_i with the input stage of its corresponding lowpass ΔΣ ADC 180_i. As a result, input multipliers 160_i do not add additional nonlinearities into the delta sigma conversion unit 501.

[0080] FIGS. 13A, 13B, 13C, 14A and 14B illustrate a particular, non-limiting example of the invention. In this example, it is assumed that the bandwidth control unit 300 of control unit 700 has decided or has been instructed to form a single broadband modulator using four channels, that is N=4. The four channels cover the entire Nyquist band of the delta sigma conversion unit 501. The signals at various points in the delta sigma conversion unit 501 are calculated to demonstrate the basic principle of delta sigma analog-to-digital conversion.

[0081] FIG. 13A illustrates an exemplary set of Hadamard codes 151 that may be produced by a Hadamard code generator 1500 for number N of ΔΣ channels 155_i in delta sigma conversion unit 501 with N=4, according to an embodiment of the present invention. Since H₁=[1], it follows that

[0082] $H_2=\left[\begin{array}{cc}{H}_1& H_1\\ H_1& -H_1\end{array}\right]=\left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right]$
and

[0083] $H_4=\left[\begin{array}{cc}{H}_2& H_2\\ H_2& -H_2\end{array}\right]=\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& -1& 1& -1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\end{array}\right].$
The Hadamard codes 151 produced by Hadamard code generator 1500 for N=4 are the rows of Hadamard matrix H₄. Hence the Hadamard codes 151 for N=4 are:
C1=[1 1 1 1]
C2=[1 −1 1 −1]
C3=[1 1 −1 −1]
C4=[1 −1 −1 1].
Hadamard input multiplier sequences 152 are generated from Hadamard codes C1, C2, C3 and C4 for input multipliers 160. Similarly, Hadamard output multiplier sequences 153 are generated from Hadamard codes C1, C2, C3 and C4 for output multipliers 240.

[0084] FIG. 13B illustrates an exemplary set of Hadamard input multiplier sequences 152 that may be used by input multipliers 160 in a delta sigma conversion unit 501 for number N of ΔΣ channels 155_i in delta sigma conversion unit 501 with N=4, according to an embodiment of the present invention. Hadamard input multiplier sequences 152 z_i(n) are obtained from Hadamard codes 151 by the following formula:
z_i(n)=C_i[n mod N]
where A mod B is the modulus function that returns the integer remainder value when A is divided by B, and i is the index for each ΔΣ channel 155 in delta sigma conversion unit 50, i running from 1 to N. Other formulas for obtaining Hadamard input multiplier sequences 152 z_i(n) are also possible and various lookup tables may also be used to supply the Hadamard code sequences as described in relation to FIGS. 4B and 4C. The first element in matrices C_i has index 0, therefore C_i elements have indices from 0 to N−1. When N=4 and signal samples n run from 0 to 15, Hadamard input multiplier sequences 152 for the second ΔΣ channel 155, i=2, are:
z₂(0)=C₂[0 mod 4]=C₂[0]=1 z₂(4)=z₂(8)=z₂(12)=z₂(0)=1
z₂(1)=C₂[1 mod 4]=C₂[1]=−1 z₂(5)=z₂(9)=z₂(13)=z₂(1)=−1
z₂(2)=C₂[2 mod 4]=C₂[2]=1 z₂(6)=z₂(10)=z₂(14)=z₂(2)=1
z₂(3)=C₂[3 mod 4]=C₂[3]=−1 z₂(7)=z₂(11)=z₂(15)=z₂(3)=−1
Hadamard input multiplier sequences 152 for all N ΔΣ channels 155 when N=4 are shown in FIG. 13B. In each ΔΣ channel 155 i, sample n of signal 5A is multiplied by z_i(n).

[0085] FIG. 13C illustrates a set of exemplary Hadamard output multiplier sequences 153 that may be used by output multipliers 240_i in a delta sigma conversion unit 501, for number N of ΔΣ channels 155_i in delta sigma conversion unit 50 with N=4, according to an embodiment of the present invention. The purpose of Hadamard output multiplier sequences 153 is to undo the frequency decomposing action of Hadamard input multiplier sequences 152. Within each ΔΣ channel 155, Hadamard output multiplier sequence 153 t_i(n) is a delayed version of Hadamard input multiplier sequence 152 z_i(n) to account for processing and signal delays between input multiplier 160 and output multiplier 240. For the specific case of Hadamard input multiplier sequences 152 in FIG. 13B, a set of Hadamard output multiplier sequences 153 t_i(n) that may be used by output multipliers 240_i can be obtained from Hadamard codes 151 by the following formula:
t_i(n)=C_i[(n+1)mod N)]
When N=4 and signal samples n run from 0 to 15, Hadamard output multiplier sequences 153 for the second ΔΣ channel 155, i=2, are:
t₂(0)=C₂[1 mod 4]=C₂[1]=−1 t₂(4)=t₂(8)=t₂(12)=t₂(0)=−1
t₂(1)=C₂[2 mod 4]=C₂[2]=1 t₂(5)=t₂(9)=t₂(13)=t₂(1)=1
t₂(2)=C₂[3 mod 4]=C₂[3]=−1 t₂(6)=t₂(10)=t₂(14)=t₂(2)=−1
t₂(3)=C₂[4 mod 4]=C₂[0]=1 t₂(7)=t₂(11)=t₂(15)=t₂(3)=1
Hadamard output multiplier sequences 153 for all ΔΣ channels 155 when N=4 are shown in FIG. 13C. In each ΔΣ channel 155 i, sample n of signal going into output multipliers 240 is multiplied by t_i(n).

[0086] FIG. 14A illustrates aspects of the operation for obtaining the output of a signal passed through Hadamard input multipliers 160_i and Hadamard output multipliers 240_i in a delta sigma conversion unit 501, according to a particular example of an embodiment of the present invention. For simplicity, the example uses a chain of Hadamard input multipliers 160 followed by Hadamard output multipliers 240, without the lowpass ΔΣ ADCs 180 in the middle. The length of identical H(z) digital correction filters 200 is chosen to be 6. The 6 samples from a signal 5A enter Hadamard input multipliers 160 S0=[x[0], x[1], x[2], x[3], x[4], x[5]]. Inside input multipliers 160, signal S0 is multiplied by corresponding Hadamard input multipliers sequences 152 labeled as C1 set, C2 set, C3 set, C4 set in FIG. 13B. Signals S1, S2, S3, S4 for channels 1,2,3 and 4 respectively, are output from input multipliers 160:
S1=[x[0]*z₁(0),x[1]*z₁(1),x[2]*z₁(2),x[3]*z₁(3),x[4]*z₁(4),x[5]*z₁(5)]=[x[0],x[1],x[2],x[3],x[4],x[5]];
S2=[x[0]*z₂(0),x[1]*z₂(1),x[2]*z₂(2),x[3]*z₂(3),x[4]*z₂(4),x[5]*z₂(5)]=[x[0],−x[1],x[2],−x[3],x[4],−x[5]];
and so on.
Signals S1, S2, S3 and S4 are sent to digital correction filters 200 or order 6. Each order 6 digital correction filter 200 is represented by filter functions [h(0),h(1),h(2),h(3),h(4),h(5)]. Signals G1, G2, G3, and G4 for channels 1,2,3 and 4 respectively, are output from digital correction filters 200:
G1=h(0)x[0]+h(1)x[1]+h(2)x[2]+h(3)x[3]+h(4)x[4]+h(5)x[5];
G2=h(0)x[0]−h(1)x[1]+h(2)x[2]−h(3)x[3]+h(4)x[4]−h(5)x[5];
and so on.
Signals G1, G2, G3 and G4 are then input into Hadamard output multipliers 240 where they are each multiplied by the Hadamard output multiplier sequences 153 corresponding to the order of the last sample input into the system, x[5]. The Hadamard output multiplier sequences 153 corresponding to the order of the last sample input into the system are labeled t₁(5) ,t₂(5),t₃(5),t₄(5) in FIG. 5C. The output obtained from network adder 260 is sum G1t₁(5)+G2t₂(5)+G3t₃(5)+G4t₄ (5). As seen in FIG. 14A, the output is 4h(2)x[2].

[0087] FIG. 14B illustrates aspects of the operation for obtaining the output of a signal passed through Hadamard input multipliers 160_i and Hadamard output multipliers 240 in a delta sigma conversion unit 501, according to a particular example of an embodiment of the present invention. The algorithm presented in FIG. 14A is repeated in FIG. 14B for the next time instant when the input is S0=[x[1], x[2], x[3], x[4], x[5], x[6]]. Within each channel, signal S0 is multiplied by corresponding Hadamard input multipliers sequences 152 labeled as D1 set, D2 set, D3 set, D4 set in FIG. 13B, sent to digital correction filters 200, multiplied by the Hadamard output multiplier sequences 153 t₁(6),t₂(6),t₃(6),t₄(6) from FIG. 5C corresponding to the order of the last sample input into the system, x[6], and added back in network adder 260. The result is 4h(2)x[3]. Therefore the output signal of network adder 260 is simply a delayed and scaled version of the input signal.

[0088] The parallel, adaptive delta sigma ADC 100 is compatible with high speed, BiCMOS mixed signal processes, SiGe processes, and monolithic integration, as well as ASIC implementation. The filtering operations inside parallel, adaptive ADC 100 are linear, and therefore can be interchanged and/or combined.

[0089] Although detailed embodiments and implementations of the present invention have been described above, it should be apparent that various modifications are possible without departing from the spirit and scope of the present invention.