Amr Elshazly

A 16mW 78dB-SNDR 10MHz-BW CT-ΔΣ ADC using residue-cancelling VCO-based quantizer

Voltage-controlled oscillator (VCO) based analog-to-digital conversion presents an attractive means of implementing high-bandwidth oversampling ADCs [1,2]. They exhibit inherent noise-shaping properties and can operate at low supply voltages and high sampling rates [1-3]. However, usage of VCO-based ADCs has been limited due to their nonlinear voltage-to-frequency (V-to-F) transfer characteristic, which severely degrades their distortion performance. Digital calibration is used to combat nonlinearity in an open-loop VCO-based ADC, but 1st-order noise-shaping mandates high OSRs, thus increasing power dissipation in digital circuits, even in a nanometer-scale CMOS process [1]. In [2], nonlinearity is suppressed by embedding the VCO in a ΔΣ loop. While this technique works in principle, the need for large loop gain at high frequencies makes it very difficult to achieve high SNDR. For instance, the suppression level near the band edge is approximately 20dB for a VCO-based 2nd-order modulator operating with an over-sampling ratio (OSR) of 30. Our ADC overcomes the impact of VCO non-linearity by minimizing the input signal processed by the VCO. The prototype achieves 78.3dB SNDR in a 10MHz signal bandwidth at 600MHz sampling rate, while consuming 16mW power.

A 16-mW 78-dB SNDR 10-MHz BW CT

This paper presents a continuous-time (CT) ΔΣ modulator using a VCO-based internal quantizer. It incorporates a nonlinear VCO as the second stage in a two-stage residue canceling quantizer (RCQ) and mitigates the impact of its nonlinearity by spanning only a small region of the VCOs V-to-F nonlinear tuning curve. The order of noise shaping is increased by placing the RCQ in a continuous-time ΔΣ loop. Using only a first order loop filter, the proposed ΔΣ modulator achieves second order noise shaping. Fabricated in a 90-nm CMOS process, the prototype modulator occupies an active area of 0.36 mm2 and consumes 16 mW power. It achieves a peak SNDR of 78.3 dB in 10-MHz bandwidth and an SFDR of better than 85 dB when clocked at 600 MHz. The figure of merit of the modulator is 120 fJ/conv-step.

\Delta \Sigma

Clock and data recovery (CDR) circuits with wide frequency acquisition range offer flexibility in optical communication networks, help reduce link power through activity-based rate adaptation, and minimize cost with a single-chip multi-standard solution. Extracting the bit rate from the incoming random data stream is the main challenge in implementing reference-less CDRs. A conventional rotational frequency detector has a limited acquisition range of about ±50% of the VCO frequency, consumes large power, and is susceptible to harmonic locking. Extending its range requires additional high-speed circuitry and a complex state machine [1]. The DLL-based architecture in [2] requires passing high-speed data through a long string of power-hungry buffers, imposes stringent matching requirements, and works only with ring oscillators. Other approaches require detailed statistical [3] or timing analysis [4]. Further, all the above techniques are only suitable for full-rate CDRs. In this paper, we present a reference-less half-rate CDR that uses a sub-harmonic extraction method to achieve unlimited frequency acquisition range. This technique is capable of locking the CDR to within 40ppm of any sub-rate of the data (making it applicable for any sub-rate CDR architecture), while being immune to undesirable harmonic locking. This CDR also integrates a calibration loop to improve robustness to input duty cycle error.

ADC Using Residue-Cancelling VCO-Based Quantizer

A highly-digital clock multiplication architecture that achieves excellent jitter and mitigates supply noise is presented. The proposed architecture utilizes a calibration-free digital multiplying delay-locked loop (MDLL) to decouple the tradeoff between time-to-digital converter (TDC) resolution and oscillator phase noise in digital phase-locked loops (PLLs). Both reduction in jitter accumulation down to sub-picosecond levels and improved supply noise rejection over conventional PLL architectures is demonstrated with low power consumption. A digital PLL that employs a 1-bit TDC and a low power regulator that seeks to improve supply noise immunity without increasing loop delay is presented and used to compare with the proposed MDLL. The prototype MDLL and DPLL chips are fabricated in a 0.13 μm CMOS technology and operate from a nominal 1.1 V supply. The proposed MDLL achieves an integrated jitter of 400 fs rms at 1.5 GHz output frequency from a 375 MHz reference clock, while consuming 890 μ W. The worst-case supply noise sensitivity of the MDLL is 20 fspp/mVpp which translates to a jitter degradation of 3.8 ps in the presence of 200 mV supply noise. The proposed clock multipliers occupy active die areas of 0.25 mm2 and 0.2 mm2 for the MDLL and DPLL, respectively.

A 0.5-to-2.5 Gb/s Reference-Less Half-Rate Digital CDR With Unlimited Frequency Acquisition Range and Improved Input Duty-Cycle Error Tolerance

A digital phase-locked loop (DPLL) employs a linear proportional path, a double integral path, bandwidth and tuning range tracking; and a novel delta-sigma digital to analog converter to achieve low jitter, wide operating range and low power. The proposed proportional path decouples the detector quantization error and oscillator noise bandwidth tradeoff and helps maximize bandwidth to suppress digitally controlled oscillator (DCO) phase noise in a power efficient manner. A double integral path alleviates the tradeoff between DCO tuning range and its frequency quantization error. The high resolution of the DCO was maintained over a wide range of sampling clock frequencies by using a delta-sigma digital to analog converter and a continuously tunable switched-RC filter. Bandwidth and tuning range tracking are employed to achieve low jitter over the entire operating range. The prototype DPLL, fabricated in a 90 nm CMOS process, operates from 0.7 GHz to 3.5 GHz. At 2.5 GHz, the proposed DPLL consumes only 1.6 mW power from a 1 V supply and achieves 1.6 ps and 11.6 ps of long-term r.m.s and peak jitter, respectively.

Clock Multiplication Techniques Using Digital Multiplying Delay-Locked Loops

A digital fractional-N PLL that employs a high resolution TDC and a truly ΔΣ fractional divider to achieve low in-band noise with a wide bandwidth is presented. The fractional divider employs a digital-to-time converter (DTC) to cancel out ΔΣ quantization noise in time domain, thus alleviating TDC dynamic range requirements. The proposed digital architecture adopts a narrow range low-power time-amplifier based TDC (TA-TDC) to achieve sub 1 ps resolution. By using TA-TDC in place of a BBPD, the limit cycle behavior that plagues BB-PLLs is greatly suppressed by the TA-TDC, thus permitting wide PLL bandwidth. The proposed architecture is also less susceptible to DTC nonlinearity and has faster settling and tracking behavior compared to a BB-PLL. Fabricated in 65 nm CMOS process, the prototype PLL achieves better than -106 dBc/Hz in-band noise and 3 MHz PLL bandwidth at 4.5 GHz output frequency using 50 MHz reference. The PLL consumes 3.7 mW and achieves better than 490 fsrms integrated jitter. This translates to a FoMJ of -240.5 dB, which is the best among the reported fractional-N PLLs.

A 0.7-to-3.5 GHz 0.6-to-2.8 mW Highly Digital Phase-Locked Loop With Bandwidth Tracking

A digital phase-locked loop (DPLL) employs noise cancellation to mitigate performance degradation due to noise on the ring oscillator supply voltage. A deterministic test signal-based digital background calibration is used to accurately set the cancellation gain and thus achieve accurate cancellation under different process, voltage, temperature, and operating frequency conditions. A hybrid, linear proportional control and bang-bang digital integral control, is used to obviate the need for a high-resolution time-to-digital converter and reduce jitter due to frequency quantization error. Fabricated in 0.13 m CMOS technology, the DPLL operates from a 1.0 V supply and achieves an operating range of 0.4-to-3 GHz. At 1.5 GHz, the DPLL consumes 2.65 mW power wherein the cancellation circuitry consumes about 280 W. The proposed noise cancellation scheme reduces the DPLLs peak-to-peak jitter from 330 to 50 ps in the presence of a 30 mV 10 MHz supply noise tone, and the DPLL peak-to-peak jitter is 50 ps in the absence of any supply noise. The DPLL occupies an active die area of 0.08 mm, of which the calibration logic and cancellation circuitry occupy only 12.5%.

A 3.7 mW Low-Noise Wide-Bandwidth 4.5 GHz Digital Fractional-N PLL Using Time Amplifier-Based TDC

A high-resolution time-to-digital converter (TDC) using switched-ring oscillators (SROs) is presented. Leveraging oversampling and noise shaping, the proposed SRO-TDC achieves high resolution without the need for calibration. Ring oscillators are switched between two frequencies to achieve noise shaping of the quantization error in an open-loop manner. By decoupling the sampling clock and input carrier frequencies, SRO-TDC is capable of operating at high oversampling ratios (OSRs). This paper also discusses different noise sources and quantization/device noise tradeoffs in noise-shaping TDCs and presents techniques to characterize TDC linearity, range, and noise performance. Fabricated in 90 nm CMOS technology, the proposed TDC operates over a wide range of input carrier frequencies (0.6-750 MHz) and sampling rates (50-750 MS/s). At 500 MS/s and 80 MHz carrier frequency, it achieves an integrated noise of 315 fs in a 1 MHz bandwidth while consuming 1.5 mW from a 1 V supply. The SRO-TDC occupies an active die area of only 0.02 mm2.

A 0.4-to-3 GHz Digital PLL With PVT Insensitive Supply Noise Cancellation Using Deterministic Background Calibration

Ever-growing demand for higher communication bandwidth in high performance compute systems is driving the need for energy-efficient multi-Gb/s I/O serial links. Improved power efficiency was demonstrated using adaptive supply regulation [1, 2]. However, power losses in the DC-DC converter needed to generate the optimal supply voltage and the difficulty in operating analog circuits at low voltages limit the power savings. Instead of scaling the supply with the data rate, we seek to operate with two fixed voltages and eliminate the need for a high-efficiency DC-DC converter. To this end, this paper presents a serial link using a highly efficient current recycling-based implicit DC-DC conversion to generate 0.6V from a 1.2V supply. Highly digital clocking circuits capable of operating at 0.6V maximize power savings. A 0.5-to-4Gb/s serial-link transceiver is designed in a 1.2V LP 90nm CMOS process to operate with a short channel and ple-siochronous timing. The transceiver dissipates 1.9mW/Gb/s at 3.2Gb/s.

A Noise-Shaping Time-to-Digital Converter Using Switched-Ring Oscillators—Analysis, Design, and Measurement Techniques

A time-to-digital converter (TDC) employs a phase-reference second-order continuous-time delta-sigma modulator to achieve high resolution and low power. The modulator operates on the phase of the input signal and generates an equivalent noise-shaped one-bit output data stream. Fabricated in an LP 90nm CMOS process, the prototype TDC achieves better than 2.4ps resolution over a 3.2ns range in a 1MHz signal bandwidth while consuming 2.1mW from a 1.2V supply.