BLUE-NOISE-MODULATED SIGMA-DELTA ANALOG-TO-DIGITAL

CONVERTER

Reference to Related Application

[0001] The present application claims the benefit of U.S. Provisional Patent Application No. 60/719,550, filed September 23, 2005, whose disclosure is hereby incorporated by reference in its entirety into the present disclosure.

Field of the Invention

[0002] The present invention is directed to a sigma-delta analog-to-digital converter and more particularly to such a converter in which the multiplying sequence has blue-noise spectral characteristics.

Description of Related Art

[0003] As CMOS technology scaling continues to reduce feature size, chip designers will integrate more and more analog and digital circuitry together on the same die in an effort to reduce cost. However, integration of systems-on-a-chip (SoC) requiring the placement of noise-sensitive analog blocks and noisy digital signal processing blocks together on a common substrate will most likely continue to increase the amount of substrate noise

generated by the digital circuitry. In particular, mixed-signal designs such as σδ (sigma-

delta) ADC's, where the analog and digital components cannot necessarily be placed far from each other, will see significant performance degradations caused by substrate noise.

The SNDR of a σδ modulator may decrease by over 20 dB in the presence of noisy

digital circuitry such as toggling inverters.

[0004] It is well known that σδ ADC's are suitable for high resolution and low-to-moderate

bandwidth applications. However, when substrate noise is introduced into the σδ modulator, dynamic range is sacrificed.

[0005] Fig. 1 shows a conventional second-order σδ modulator 100. In the modulator 100, a

signal received through an input 102 is applied to a subtracting circuit 104, where a feedback signal (to be described later) is subtracted from the signal. The resulting signal is integrated in a non-delayed integrator 106. The integrated signal is applied to a subtracting circuit 108, where the feedback signal is subtracted from the integrated signal. The resulting signal is integrated in a non-delayed integrator 110 and a delay circuit 112, which together form a sample-delayed integrator 114. The additive noise contribution from the modulator's quantizer is modeled as a signal E _Q received through an input 116 and added to the integrated signal in an adding circuit 118. The output signal is applied to an output 120 as well as to a feedback loop 122, where it is applied as the above- mentioned feedback signal to the subtracting circuits 104 and 108. The conventional modulator 100 does not adequately address the noise problem.

[0006] Similar work is disclosed in United States Patents 6,707,409 and 7,038,532, whose disclosures are hereby incorporated by reference in their entireties into the present disclosure. However, those patents do not directly address the noise problem in the same manner as the present invention.

Summary of the Invention

[0007] It is therefore an object of the invention to provide a σδ ADC which overcomes the

above difficulties of the prior art.

[0008] To achieve the above and other objects, the present invention is directed to a σδ ADC

using blue-noise modulation (or, more generally, modulation with a random or pseudorandom noise sequence) to reduce the effect of the substrate noise. The blue-noise

modulated σδ ADC is obtained by placing pairs of blue-noise multipliers before and after

each non-delayed integrator in a σδ ADC such as the conventional σδ ADC shown in

Fig. 1. In the case of a sample-delayed integrator, the integrator is first separated into a non-delayed integrator followed by the delay element. The multiplying sequence, e[n],

used to modulate the signals throughout the blue-noise σδ modulator is a sequence of l 's

and -1 's that has blue-noise spectral characteristics (i.e., the spectrum has a low frequency deficiency).

[0009] The σδ modulator architecture presented herein utilizes blue-noise modulation,

whereby a signal with blue-noise spectral characteristics, used as a chopper signal, eliminates the substrate noise spectral peaks by spreading them across the entire ADC spectrum. The noise is then shaped away from the baseband input signal to high frequency as is typical for oversampling σδ ADC's. The present invention eliminates the strong distortion component near DC created by harmonics of the digital sampling clock.

[0010] The present invention can be implemented as an easy modification of mature conventional technology. Simulations of such architecture have proved the ability to suppress substrate noise generated by noisy digital blocks placed on the same die as sensitive analog circuits. The result is in an overall improvement in SNR of over 14 dB.

[0011] In addition to reducing the effects of substrate noise, the present invention also eliminates the effects of integrator op-amp non-idealities, such as 1/f noise and DC offset,

and DAC DC offset and even-order non-linearities. An example utilizing a second-order

blue-noise modulated σδ ADC with a 1-bit quantizer is presented as an illustrative rather

than limiting embodiment. Additionally, a method to generate the required blue-noise sequence is presented. Simulation results demonstrate that this architecture achieves a 14

dB improvement in SNR over the traditional second-order σδ ADC.

Brief Description of the Drawings

[0012] A preferred embodiment of the present invention will be set forth in detail with reference to the drawings, in which:

[0013] Fig. 1 is a block diagram showing a conventional second-order σδ modulator;

[0014] Figs. 2A-2C are block diagrams showing steps in the design of a blue-noise-

modulated σδ modulator according to the preferred embodiment;

[0015] Figs. 3A and 3B are plots of input signal spectra before and after blue-noise modulation, respectively;

[0016] Fig. 4 is a plot showing a representative blue-noise spdctrum;

[0017] Fig. 5 is a block diagram showing an all-digital σδ modulator for generating the blue-

noise sequence;

[0018] Fig. 6 is a plot showing an output spectrum of a conventional second-order, one-bit

σδ modulator; and

[0019] Fig. 7 is a plot showing an output spectrum of a blue-noise-modulated σδ modulator

according to the preferred embodiment.

Detailed Description of the Preferred Embodiment

[0020] A preferred embodiment of the present invention will be set forth in detail with reference to the drawings, in which like reference numerals refer to like elements throughout.

[0021] Figs. 2A-2C show steps in the design of the σδ ADC of the preferred embodiment,

using as a starting point the conventional ADC of Fig. 1.

[0022] As shown in Fig. 2A, the blue-noise sequences 202 are introduced into the modulator structure in pairs to form a structure 200A. Thus, they effectively perform multiplication by 1 , leaving the overall modulator function unchanged. Once the blue-noise multipliers 202 are inserted, as shown in Fig. 2A, one multiplier 202 from each pair is moved

through the modulator to produce the blue-noise modulated σδ modulator structure 200C

shown in Fig. 2C. Fig. 2B shows the intermediate stage 200B in rearranging the blue-

noise multipliers to create the blue-noise modulated σδ ADC. The dashed arrows in Figs.

2A and 2B depict the direction in which a particular blue-noise multiplier is moved. This

structural transformation can be applied to any traditional σδ modulator to create a blue-

noise modulated σδ modulator. The architecture 200C presented in Fig. 2C incorporates a

non-delayed blue-noise modulated (BNM) integrator structure 204 and a sample-delayed BNM integrator structure 206. As noted above, E _Q models the additive noise contribution

from the modulator's quantizer. A distinct advantage of the blue-noise modulated σδ

modulator 200C is that it can be easily implemented, based on mature σδ modulator technology.

[0023] The input signal can be modulated with the same blue-noise (random, pseudorandom) sequence prior to being connected to the subtracting circuit, as shown in Fig. 2C. The sequence used to modulate the input signal can be a delayed or advanced version of the sequence used with in the σδ modulator.

[0024] In addition to spreading substrate noise, which exhibits strong spectral peaks, blue- noise modulation also negates the effects of integrator op-amp and DAC non-idealities (l//noise, DC offset, and even-order nonlinearities). This is due to the fact that blue-noise modulation shapes the input signal about the Nyquist frequency before it sees the non- idealities from the integrator op-amps and DAC. The input signal spectra before and after blue-noise modulation are shown in Figs. 3A and 3B, respectively. The undesirable effects will remain at low frequency where the quantization noise is shaped by the blue- noise modulated integrator structures. When the output blue-noise modulation is performed, the input signal is returned to baseband, while the shaped quantization noise along with the low-frequency 1 // ^" noise and DC offsets are pushed toward high frequency.

[0025] The spectrum of a blue-noise sequence consists of low frequency deficiencies and uncorrelated high-frequency fluctuations that are classified as high-frequency white noise. The sharp transition between the energy-limited low frequency portion of the blue-noise spectrum and the high-frequency white noise occurs at the principal frequency. The principal frequency for a particular blue-noise sequence is commonly denoted f _g . The spectrum of a typical blue-noise sequence is shown in Fig. 4.

[0026] The binary blue-noise sequence necessary for blue-noise modulation can be generated

using an all-digital σδ modulator. Replacing the integrators from traditional σδ

modulators with accumulators creates an all-digital σδ architecture which is shown in

Fig. 5 as 500. In the architecture 500, an adder 502 adds a feedback signal, to be explained below, to the contents of a B+l register 504, which samples the output of the adder 502 at a sampling frequency /J. The result of the adder 502 is also supplied to a subtracting circuit 506, where the feedback signal is subtracted. The result is supplied to another adder 508, where a second feedback signal from a feedback loop 510 is added. The addition result is supplied to a B+2 register 512, which samples it at the same

sampling frequency f _s . The output of the B+2 register 512 is supplied to both the feedback loop 510 and an adder 514 to which a dither is added. The result is supplied to a quantizer 516, where e[n] is derived, to supply an output signal of B bits. The output signal is supplied to both an output 518 and a feedback loop 520, where it undergoes digital code conversion in a digital code converter 522 to supply the feedback signal which is supplied to the adder 520 and the subtracting circuit 506. Since the blue-noise

sequences used in the blue-noise modulated σδ modulator consist only of 1 's and -1 's,

the quantizer needs to resolve B = I bit. Similar algorithms that produce blue-noise sequences have been proposed for fractional-N PLL applications and for DAC dynamic-

element matching in multi-bit σδ ADC applications.

[0027] Alternatively, the sequence e[n] can be stored, entirely or in part, in a local memory shown in Fig. 2C as 208.

[0028] The behavioral simulation results presented below were obtained from

Matlab/Simulink models of a second-order, 1-bit σδ modulator. The input used for the

simulations was a sinusoidal signal with frequency β _n - 4.1 kHz lying within an 11.025 kHz bandwidth. The OSR was selected to be 512, resulting in a sampling frequency off _s — 11.2896 MHz. The substrate coupling noise used for the simulations was obtained in the

lab from Maxim ICs MAXl 403 18-bit σδ ADC. The measured noise, which contained

the digital sampling clock and several strong higher harmonics and subharmonics, was

imported into Matlab for the simulations and injected into the first integrator in the σδ

modulator. Also, in order to provide realistic simulations, non-ideal op-amp models were used for the integrators. The simulations were performed assuming room temperature, a finite op-amp gain, finite gain-bandwidth, slew rate, and saturation. In addition to the op- amp non-idealities, clock jitter was also taken into account.

[0029] The output spectrum of a typical second-order 1-bit σδ modulator is shown in Fig. 6.

The output spectrum clearly depicts the modulator's vulnerability to low-frequency noise, which limits the SNR to 88 dB in simulation. The output spectrum of the blue-noise

modulated σδ modulator, shown in Fig. 7, reveals that this architecture suppresses the

low-frequency distortions present in the traditional modulator spectrum. The resulting SNR for the blue-noise modulated architecture using the same parameters as the simulation without blue-noise modulation is 102 dB. The proposed architecture

demonstrates a 14 dB improvement in SNR from traditional σδ modulator designs, which

results in an increase in effective resolution of over 2 bits.

[0030] While a preferred embodiment has been described in detail, those skilled in the art who have reviewed the present disclosure will readily appreciate that other embodiments

can be realized within the scope of the invention. For example, a σδ ADC of any order can be implemented. Also, any suitable source of a binary blue-noise sequence can be used. Moreover, any feature disclosed in U.S. Patent No. 6,707,409 or 7,038,532 can be incorporated into the present invention as needed. Therefore, the invention should be construed as limited only by the appended claims.