Smart and high-performance digital-to-analog converters with dynamic-mismatch mapping

Y. Tang

Research output: ThesisPhd Thesis 1 (Research TU/e / Graduation TU/e)

2483 Downloads (Pure)

Abstract

The trends of advanced communication systems, such as the high data rate in multi-channel base-stations and digital IF conversion in software-defined radios, have caused a continuously increasing demand for high performance interface circuits between the analog and the digital domain. A Digital-to-Analog converter (DAC) is such an interface circuit in the transmitter path. High bandwidth, high linearity and low noise are the main design challenges in high performance DACs. Current-steering is the most suitable architecture to meet these performance requirements. The aim of this thesis is to develop design techniques for high-speed high-performance Nyquist current-steering DACs, especially for the design of DACs with high dynamic performance, e.g. high linearity and low noise. The thesis starts with an introduction to DACs in chapter 2. The function in time/frequency domain, performance specifications, architectures and physical implementations of DACs are brie y discussed. Benchmarks of state-of-the-art published Nyquist DACs are also given. Chapter 3 analyzes performance limitations by various error sources in Nyquist current-steering DACs. The outcome shows that in the frequency range of DC to hundreds of MHz, mismatch errors, i.e. amplitude and timing errors, dominate the DAC linearity. Moreover, as frequencies increase, the effect of timing errors becomes more and more dominant over that of amplitude errors. Two new parameters, i.e. dynamic-INL and dynamic-DNL, are proposed to evaluate the matching of current cells. Compared to the traditional static-INL/DNL, the dynamic-INL/DNL can describe the matching between current cells more accurately and completely. By reducing the dynamic-INL/DNL, the non-linearities caused by all mismatch errors can be reduced. Therefore, both the DAC static and dynamic performance can be improved. The dynamic-INL/DNL are frequency-dependent parameters based on the measurement modulation frequency fm. This fm determines the weight between amplitude and timing errors in the dynamic-INL/DNL. Actually, this gives a freedom to optimize the DAC performance for different applications, e.g. low fm for low frequency applications and high fm for high frequency applications. Chapter 4 summarizes the existing design techniques for intrinsic and smart DACs. Due to technology limitations, it is diffcult to reduce the mismatch errors just by intrinsic DAC design with reasonable chip area and power consumption. Therefore, calibration techniques are required. An intrinsic DAC with calibration is called a smart DAC. Existing analog calibration techniques mainly focus on current source calibration, so that the amplitude error can be reduced. Dynamic element matching is a kind of digital calibration technique. It can reduce the non-linearities caused by all mismatch errors, but at the cost of an increased noise oor. Mapping is another kind of digital calibration technique and will not increase the noise. Mapping, as a highly digitized calibration technique, has many advantages. Since it corrects the error effects in the digital domain, the DAC analog core can be made clean and compact, which reduces the parasitics and the interference generated in the analog part. Traditional mapping is static-mismatch mapping, i.e. mapping only for amplitude errors, which many publications have already addressed on. Several concepts have also been proposed on mapping for timing errors. However, just mapping for amplitude or timing error is not enough to guarantee a good performance. This work focuses on developing mapping techniques which can correct both amplitude and timing errors at the same time. Chapter 5 introduces a novel mapping technique, called dynamic-mismatch mapping (DMM). By modulating current cells as square-wave outputs and measuring the dynamic-mismatch errors as vectors, DMM optimizes the switching sequence of current cells based on dynamic-mismatch error cancelation such that the dynamic-INL can be reduced. After reducing the dynamic-INL, the non-linearities caused by both amplitude and timing errors can be significantly reduced in the whole Nyquist band, which is confirmed by Matlab behavioral-level Monte-Carlo simulations. Compared to traditional static-mismatch mapping (SMM), DMM can reduce the non-linearities caused by both amplitude and timing errors. Compared to dynamic element matching (DEM), DMM does not increase the noise floor. The dynamic-mismatch error has to be accurately measured in order to gain the maximal benefit from DMM. An on-chip dynamic-mismatch error sensor based on a zero-IF receiver is proposed in chapter 6. This sensor is especially designed for low 1/f noise since the signal is directly down-converted to DC. Its signal transfer function and noise analysis are also given and con??rmed by transistor-level simulations. Chapter 7 gives a design example of a 14-bit current-steering DAC in 0.14mum CMOS technology. The DAC can be configured in an intrinsic-DAC mode or a smart-DAC mode. In the intrinsic-DAC mode, the 14-bit 650MS/s intrinsic DAC core achieves a performance of SFDR>65dBc across the whole 325MHz Nyquist band. In the smart-DAC mode, compared to the intrinsic DAC performance, DMM improves the DAC performance in the whole Nyquist band, providing at least 5dB linearity improvement at 200MS/s and without increasing the noise oor. This 14-bit 200MS/s smart DAC with DMM achieves a performance of SFDR>78dBc, IM3
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Electrical Engineering
Supervisors/Advisors
  • van Roermund, Arthur, Promotor
  • Hegt, Hans, Copromotor
Award date6 Oct 2010
Place of PublicationEindhoven
Publisher
Print ISBNs978-90-386-2322-1
DOIs
Publication statusPublished - 2010

Fingerprint

Dive into the research topics of 'Smart and high-performance digital-to-analog converters with dynamic-mismatch mapping'. Together they form a unique fingerprint.

Cite this