On the Design of High-Performance mm-Wave CMOS VCOs by Milad Haghi Kashani B. Sc., Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran, Aug 2016 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (ELECTRICAL AND COMPUTER ENGINEERING) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) August 2018 © Milad Haghi Kashani, 2018 ii The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, a thesis/dissertation entitled: On the Design of High-Performance mm-Wave CMOS VCOs submitted by Milad Haghi Kashani in partial fulfillment of the requirements for the degree of Master of Applied Science in Electrical and Computer Engineering Examining Committee: Shahriar Mirabbasi, Electrical and Computer Engineering Supervisor Sudip Shekhar, Electrical and Computer Engineering Supervisory Committee Member Mieszko Lis, Electrical and Computer Engineering Supervisory Committee Member iii Abstract The increasing demand for higher data rates in wireless communication such as IEEE 802.15.3c wireless personal area networks (WPANs), wireless local area networks (WLANs), and next generation wireless systems, have made the development of radio-frequency (RF) and millimeter-wave (mm-wave) frequency integrated circuits (ICs) that operate in few tens of gigahertz more popular. In the context of high data rate wireless systems, the 5th generation (5G) of wireless systems is expanding its operation frequency to millimeter wave (mm-wave) bands where more signal bandwidth are available. To implement such circuits, the preferred technology is CMOS (complementary metal-oxide semiconductor) since compared to other technologies, it offers lower supply voltage, lower cost, higher levels of integration, and potentially lower power. Many CMOS solutions for RF and mm-wave applications have been proposed over the past few years and thanks to continuous reductions in feature size of CMOS devices into deep-submicron range, many efficient and high performance state-of-the-art RF and mm-wave receivers have been reported. However, efficient system integration at mm-wave frequencies in CMOS is still a challenging task. Voltage-controlled oscillators (VCOs) are indispensable in the operations of fully integrated transceiver architectures. In many cases, quadrature local oscillator (LO) signals are required for frequency conversion. Major challenges in particular in portable applications include the generation of such quadrature phase outputs with low power consumption. This also applies to emerging mm-wave applications, where one would like to achieve a wide tuning range while maintaining low-power consumption and low phase noise. In this work, we investigate several design techniques for achieving a high performance wireless receiver building block with a specific focus on LC-based VCOs. Based on the measurement results, one of the VCOs achieves an average PN of −111.9 dBc/Hz at 10 MHz offset over the entire frequency tuning range, and a TR of ~18%, from 50.1 to 59.8 GHz, resulting in a figure of merit incorporating the tuning range (FOMT) of –184 dBc/Hz. The VCO core consumes 6.2 mW from a 1-V supply and excluding the pads occupies a compact silicon area of 0.06 mm2. iv Lay Summary This thesis summarizes our efforts to further improve the performance of mm-wave signal sources that are implemented in Complementary Metal-Oxide-Semiconductor (CMOS) technology so they can achieve lower phase noise, wider tuning range, lower power consumption, and more compact chip area. The proof-of-concept prototypes are implemented in a cost-effective CMOS technology to facilitate the mm-wave operation (28 GHz and 60 GHz) for the fifth generation (5G) of communication systems and other high data-rate wireless systems, as well as medical, security and internet-of-things (IoT) applications. We have designed and implemented the proposed integrated oscillators using a 65-nm CMOS process and have experimentally validated the performance of the circuits. v Preface I, Milad Haghi Kashani, am the principal contributor of all chapters of this dissertation. Professor Shahriar Mirabbasi has supervised the research by providing financial support, technical consultation, and editing assistance on the manuscript. During my Masters studies, we have collaborated with a number of scholars and research groups as reflected in publication list below. I confirm that I have completed all the research in the following papers by myself. I just would like to thank Prof. Ehsan Afshari, Dr. Reza Molavi, Dr. Amirahmad Tarkeshdouz, and Dr. Hormoz Djahanshahi for their technical discussion and help in writing and revising the manuscripts. Please find below the list of the published works which some of them have been used as the main chapters of this thesis: Conferences: [1] M. H. Kashani, R. Molavi and S. Mirabbasi, “A Wide-Tunning-Range Low-Phase-Noise Colpitts Oscillator with Variable Capacitive Feedback,” IEEE International Symposium on Circuits and Systems (ISCAS), Florence, Italy, May 2018, pp. 1-5. [2] M. H. Kashani, R. Molavi, H. Djahanshahi and S. Mirabbasi, “On the Design of Vertical-Turn Solenoids for Magnetically Isolated Densely Integrated LC Oscillators,” IEEE International Symposium on Circuits and Systems (ISCAS), Florence, Italy, May 2018, pp. 1-5. [3] M. H. Kashani, A. Tarkeshdouz, E. Afshari, and S. Mirabbasi, “A 60-GHz CMOS Down-Conversion Mixer with High Conversion Gain and Low Noise Figure,” 16th IEEE International NEWCAS Conference, Montreal, Canada, June 2018, pp. 1-4. Recipient of Regroupement Stratégique en Microsystème du Québec (ReSMiQ) Best Paper Award. Journals: [4] M. H. Kashani, A. Tarkeshdouz, R. Molavi, E. Afshari, and S. Mirabbasi, "A Wide-Tuning-Range Low-Phase-Noise mm-Wave CMOS VCO with Switchable Transformer-based Tank," IEEE Solid-State Circuits Letters, vol. 1, no. 3, pp. 66-69, June 2018. vi Table of Contents Abstract ......................................................................................................................................... iii Lay Summary ............................................................................................................................... iv Preface ............................................................................................................................................ v Table of Contents ......................................................................................................................... vi List of Tables ............................................................................................................................... vii List of Figures ............................................................................................................................. viii List of Abbreviations .................................................................................................................... x Acknowledgements ...................................................................................................................... xi Chapter 1. Introduction ....................................................................................................... 1 1.1. Motivation ................................................................................................................................... 1 1.2. Organization ................................................................................................................................ 2 1.2.1. mm-Wave Signal Generation in CMOS Technology ............................................................................ 3 1.3. Thesis Outline .............................................................................................................................. 4 Chapter 2. A Low Phase-Noise Wide-Tuning-Range mm-wave CMOS VCO. .............. 5 2.1. Introduction and Overview of Recently Published Works .......................................................... 5 2.2. Proposed Inductive Tuning Method [12] ..................................................................................... 7 2.3. Summary of The Design Procedure ........................................................................................... 19 2.4. VCO Design And Implementation ............................................................................................ 22 2.5. Verification Of The Proposed Model ........................................................................................ 27 2.6. Discussion On The Pulling Effect Of The Proposed VCO ........................................................ 29 Chapter 3. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO .......................... 31 3.1. Introduction and Overview of Recently Published Works ........................................................ 31 3.2. The Proposed Architecture ........................................................................................................ 33 3.2.1. The Proposed Variable Capacitive Feedback Colpitts [23] ................................................................. 33 3.2.2. The Proposed Gm-boosting Technique ................................................................................................. 40 3.2.3. The Proposed Body Bias Technique .................................................................................................... 46 3.2.4. VCO Design and Implementation ........................................................................................................ 50 Chapter 4. Measurement Results...................................................................................... 54 4.1. Low-Phase-Noise Wide-Tuning-Range Mm-wave CMOS VCO ............................................. 54 4.2. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO ....................................................... 58 Chapter 5. Conclusions ...................................................................................................... 63 5.1. Low-Phase-Noise Wide-Tuning-Range Mm-wave CMOS VCO ............................................. 63 5.2. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO ....................................................... 64 References .................................................................................................................................... 66 vii List of Tables Table 1. The simulated performance summary of the proposed VCO over the process corner...65 Table 2. The performance summary of the fabricated VCO for different supply voltages……..74 Table 3. Performance summary of 60-GHz VCOs in 65-nm CMOS…………………………...75 Table 4. Performance comparison with that of the state-of-the-art designs…………………….76 viii List of Figures Figure 1. The simplified model for the proposed inductive tuning technique; (a) SW1 and SW2 are off; (b) SW1 and SW2 are on; (c) SW1 is on and SW2 is off; (d) SW1 is off and SW2 is on. 8 Figure 2. Simplified schematic of the proposed transformer. ......................................................... 9 Figure 3. Plotted C1 and C2 versus k23 with L2=210 pH, L3=250 pH, and M=115 pH. ............ 14 Figure 4. Simplified π-model of the transformer when; (a) K23=0. (b) K23=1. .......................... 15 Figure 5. Simplified model of the transformer where both switches are on (or off). ................... 15 Figure 6. Simplified pi-model for one of transformers. ................................................................ 16 Figure 7. Simulated C1 versus different values of L2 and L3. ....................................................... 18 Figure 8. Simulated C2 versus different values of L2 and L3. ....................................................... 18 Figure 9. Dependency of the coupling factor on L3. ..................................................................... 19 Figure 10. The simulated Q-factor over all switch states. ............................................................ 20 Figure 11. Simulated effective inductance over the whole frequency tuning range. .................... 20 Figure 12. Schematic of the proposed (a) VCO. (b) The proposed switch architecture. .............. 23 Figure 13. Cross section view of the proposed switchable transformer. ...................................... 24 Figure 14. Top view of the proposed switchable transformer. ..................................................... 25 Figure 15. Simulated coupling factor (k23) for different configurations. .................................... 26 Figure 16. Simulated quality factor for different configurations. ................................................. 26 Figure 17. A high frequency circuit model of an inductor. .......................................................... 28 Figure 18. Q factor comparison between that of the proposed model and transformer................ 29 Figure 19. Simulated output signal after applying a pulse at the gate of the switch. ................... 30 Figure 20. Transient response of the oscillation frequency after applying a pulse at the gate of the switch. ........................................................................................................................................... 30 Figure 21. Schematic of the Colpitts VCO with (a) Conventional (b) Proposed capacitor bank. 34 Figure 22. Half-circuit and small-signal models of : (a, c) conventional switching network (b, d) proposed switching network. ........................................................................................................ 36 Figure 23. Schematic of the conventional differential Colpitts VCO with a one-branch capacitor bank (a) conventional; (b) proposed. ............................................................................................ 39 Figure 24.The simulated loss of the tank in conventional and proposed structures. .................... 40 Figure 25. Schematic of the (a) conventional Colpitts VCO; (b) proposed gm-boosted Colpitts VCO. ............................................................................................................................................. 41 Figure 26. The simplified and small signal model of the (a, c) conventional Colpitts VCO; (b, d) proposed Colpitts VCO. ................................................................................................................ 42 Figure 27. Comparison of the simulated small-signal negative transconductance in conventional and proposed topology. ................................................................................................................. 44 Figure 28. Comparison of simulated equivalent parallel resistance in Conventional and proposed technique. ...................................................................................................................................... 45 Figure 29. Comparison of simulated equivalent parallel transconductance in Conventional and proposed technique. ...................................................................................................................... 45 Figure 30. System-level model of the proposed gm-boosting technique. (b) Simulated transient signals at V1 and V2. .................................................................................................................... 46 Figure 31. The proposed body bias technique; (a) schematic (b) corresponding signals. ............ 47 Figure 32. Comparison of the simulated negative transconductance in conventional and proposed body bias technique....................................................................................................................... 48 Figure 33. Simulated transient response of the proposed body bias technique. ........................... 48 ix Figure 34. The simulated transient drain current of M1. .............................................................. 49 Figure 35. Schematic of the fabricated VCO. ............................................................................... 51 Figure 36. Different views of the proposed tank; (a) top; b) cross sectional. ............................... 52 Figure 37. Micrograph of the fabricated 60-GHz VCO. ............................................................... 55 Figure 38. Output spectrum of the fabricated VCO. ..................................................................... 55 Figure 39. Measured phase noise at the center frequency of oscillation. ..................................... 56 Figure 40. The measured PN at 10-MHz offset over the whole frequency tuning range. ............ 57 Figure 41. The measured output power over the whole frequency tuning range. ........................ 57 Figure 42. Measured frequency tuning range over different control voltages. ............................. 58 Figure 43. Micrograph of the fabricated 28-GHz VCO. ............................................................... 59 Figure 44. Measured output spectrum of the fabricated VCO. ..................................................... 60 Figure 45. Measured phase noise of the fabricated VCO at the center frequency of oscillation.. 60 Figure 46. Measured tuning range of the fabricated VCO. ........................................................... 61 Figure 47. Measured PN and FOM at 10-MHz offset from the carrier over the entire tuning range. ............................................................................................................................................. 61 x List of Abbreviations CMOS Complementary Metal-Oxide Semiconductor CPW Coplanar Wave Guide FOM Figure of Merit FTR Frequency Tuning Range GCPW Grounded Coplanar Wave Guide gm Transconductance IF IC Intermediate Frequency Integrated Circuit LO Local Oscillator PN Phase Noise THz Terahertz TR Tuning Range VCO Voltage Control Oscillator VGS Gate-to-Source voltage VOV Overdrive Voltage Vth Threshold Voltage xi Acknowledgements First and foremost, I wish to express my special gratitude to my supervisor, Professor Shahriar Mirabbasi, for providing me with direction and technical support. I appreciate his countless advice on both research as well as on my career. I am grateful for his inspiration, encouragement and continuous support, throughout my master studies. I learnt a lot from him and our collaborations resulted in number of publications. I would like to thank Prof. Ehsan Afshari (University of Michigan Ann Arbor), Dr. Hormoz Djahanshahi (Microsemi Corporation), Dr. Reza Molavi (D-Wave Systems Inc., Canada), and Dr. Amirhossein Masnadi Shirazi (Microsemi Corporation) for fruitful discussions. I would also like to thank Canadian Microelectronics Corporation (CMC Microsystems) for providing CAD tools support and facilitating chip fabrication. I would also like to acknowledge the Natural Sciences and Engineering Research Council of Canada (NSERC) and Intel for funding this project. I would like to thank Professor S. Savafi-Naeini and Dr. Aidin Taeb at University of Waterloo and James Dietrich at University of Manitoba for providing access to measurement equipment and test setups. I also thank Dr. Roberto Rosales and Roozbeh Mehrabadi (UBC) for measurement and CAD tool support. Finally, this work is dedicated to my wonderful parents for all their love and support. I always felt them near me, even though they were living far away during my period of study in Canada. They have been my main motivation throughout my studies. Also I would like to thank my dear brother, Dr. Masoud Haghi Kashani, for paving the way for me and enormously helping me to finish my Masters successfully. xii TO MY FAMILY Fin Garden, Kashan , Iran 1 Chapter 1. Introduction 1.1. Motivation The surge in demand for high data-rate communication in wireless standards such as IEEE 802.15.3c wireless personal area networks (WPANs) and next generation wireless systems has attracted considerable attention toward development of integrated circuits and systems that operate in mm-wave frequencies [1]–[4]. In particular, the availability of an increased signal bandwidth has been the major drive for the 5th generation (5G) of wireless systems to extend their operation frequency to mm-wave bands. Although the frequency band for 5th-generation wireless systems has not been completely specified yet, the unlicensed 28-GHz and 60-GHz frequency bands are amongst the most promising candidates for such systems in many countries [5]. The millimeter-wave frequency range is prominent and necessary for basic and fundamental applications as it consists molecular vibration frequencies and absorption bands of the atmosphere and common materials. This frequency range is also useful for technological applications, such as imagers, radars, security, and medical applications. Imaging at these frequencies can be used to detect concealed weapons or explosives and reveal features in skin, teeth, and solid organs. Millimeter-wave imagers are considered as a superior alternative as compared to traditional metal detectors, since the reflection pattern of metals, plastics, ceramics and liquids are detectable quickly for radiation at these frequencies [6-8]. The function of radars is to measure the distance or speed of aircrafts, ships, automotive vehicles, and many other moving objects. Hence, the security domain facilitates one of the major areas for millimeter-waves imaging systems [3]. Another advantage of mm-wave ICs is the possibility of integrating the antenna. Because of the short wave length any interconnection with the antenna will constitute a huge parasitic element. This may cause the circuit to work improperly while it can be addressed by designing on-chip 2 antenna and transmission lines. Although on-chip antennas have been designed and fabricated before, they have not been introduced commercially. To implement such circuits, the preferred technology is CMOS (complementary metal-oxide semiconductor) since as compared to other technologies, it offers low supply voltage, low power, low cost, and high levels of integration (with possibility of integrating analog and digital functionality on the same chip) [4]. Research during the past decades provided CMOS solutions to RF and mm-wave applications. For decades, CMOS technology has dominated digital and mixed-signal ICs because of its good switching performance and low cost. CMOS technology also takes the advantages for RF and millimeter-wave circuits and systems. In addition, high-frequency circuits can be digitally controlled and calibrated [9]. Despite great advances, there are still existing bottlenecks in CMOS designs as they need to be addressed. In other words, many challenges and limitations to CMOS technology should still be alleviated to realize millimeter-wave circuits and systems. The cut-off frequencies (fT , fmax) are relatively lower than for other technologies [6]. The low breakdown voltages limits the supply voltage and AC voltage swings for reliability [2-5]. In addition, passive devices suffer from high losses and low quality factors. As the technology is scaled down, the thickness and the elevation of the metals also shrink, which increases the losses in inductors and transmission lines. In this work, we are doing our best to address those challenges and unsolved problems and push the boundaries of cost-effective electronics further toward higher speed data-rates and lower power consumption. 1.2. Organization With abovementioned motivations, the objective of this research is to study and design the following building blocks in a cost-effective CMOS process that can be used in high-performance mm-wave communication systems such as 5G systems: 3 1) Low-Phase-Noise Wide-Tuning-Range 60-GHz CMOS VCO (Chapter 2) 2) Low-Power Wide-Tuning-Range 28-GHz CMOS VCO (Chapter 3) As a proof-of-concept, for each building block, a test chip is designed, fabricated, successfully measured, and its performance is compared with the related state-of-the-art designs. The CMOS process that is used is 65-nm which provides a good cost-performance trade-off for designs that are operation in the abovementioned frequency bands. 1.2.1. mm-Wave Signal Generation in CMOS Technology The development of the integrated circuits (ICs) operating at millimeter-wave (mm-wave) frequencies has become very popular over the past few years. This popularity is the corollary of the increasing demand for high data-rate communication standards such as IEEE 802.15.3c wireless personal area network (WPAN). In addition, the 5th generation of wireless systems (5G) has extended its operation frequency to mm-wave band, where an increased signal bandwidth is available [1]. Voltage-controlled oscillators (VCOs) are amongst the essential building blocks required for operation of integrated mm-wave transceivers. One of the challenges associated with designing mm-wave VCOs is to generate a wide-band, low-phase-noise signal while maintaining low power consumption. With recently commercialized 5G cellular network, researches are looking for mm-wave and more specifically 60-GHz frequency bands, aiming for up to 10 Gb/s for wireless communication [8-10]. The signal integrity in general and the phase noise (PN) performance of the receiver or transmitter clock in particular is necessary for any high performance communication system. Most of previous researches and analytical calculations on phase noise have been focused on low RF frequency bands and in some cases they make certain assumptions which are not valid at mm-4 wave bands [11]. As an illustration, the effect of non-linear transistor capacitors are neglected. Although these capacitances have just second-order effects on the noise performance at low frequencies, at mm-wave frequencies, any small variation in these capacitances can directly or indirectly impact PN. Signal generation at mm-wave frequencies is a major challenge in solid-state electronics due to the limited transit frequency (which is the frequency at which the current gain is unity), fT, maximum oscillation frequency (which is the frequency at which the power gain is unity), fmax, breakdown voltage of active devices and the lower quality factor of passive components resulted by ohmic and substrate loss [6]. 1.3. Thesis Outline This thesis is organized as follows: 1. Investigation and implementation of low-phase-noise wide-tuning-range mm-wave oscillator design are presented in Chapter 2. 2. Investigation and implementation of low-power wide-tuning-range mm-wave oscillator design are presented in Chapter 3. 3. Measurement results of the fabricated prototypes are presented in Chapter 4. 4. Conclusion remarks are presented in Chapter 5. 5 Chapter 2. A Low Phase-Noise Wide-Tuning-Range mm-wave CMOS VCO. 2.1. Introduction and Overview of Recently Published Works The availability of wide bandwidth in the mm-wave portion of the frequency spectrum makes these bands attractive for high-data-rate applications such as wireless high-definition video streaming and medical imaging [2]–[4]. One of the main challenges in most communication systems operating in the 60 GHz and higher frequency bands is to synthesize an on-chip local oscillator (LO) with a high spectral purity and a large tuning range. Several techniques have been reported to improve the tuning range of different mm-wave VCO topologies [7][8][9]. Capacitive tuning methods are not suitable at mm-wave frequencies due to the degraded quality factor (Q) of capacitors [7]. Alternatively, inductive tuning methods have been introduced to overcome the aforementioned challenges [7][8][9]. An inductor loaded transformer is designed in [7] with switches placed at different locations on the secondary coil to increase the number of sub-bands, and the resulting tuning range. However, the relatively large number of switches required for symmetric implementation increases the contributed parasitic capacitance. Furthermore, the parasitics of off switches on the secondary coil still affects the overall performance. In [8], switchable coupled VCO cores are designed to realize a dual mode operation at the cost of a higher power consumption and larger area; the design also suffers from a low output power. In [9], switched inductors are utilized to change the effective inductance of the tank, thereby extending the tuning range at the cost of a relatively large area and high power consumption. In [4] the size of the varactor is reduced to improve the tank quality factor and the VCO oscillation frequency is varied by the drain capacitance of the MOS device facilitated by using the body effect. 6 However, this design suffers from a narrow (below 5%) frequency tuning range due to the small size of varactor. A large FTR (~28%) is achieved in [10] by utilizing dual-mode LC-tank. However, the figure-of-merit (FOM) and phase noise performance is degraded due to the use of a complex inductor structure and in turn sacrificed tank quality factor which deteriorates the moise performance. Another interesting technique is to increase the fmax and the transconductance of the cross-coupled pair by using inductive peaking at the gate of MOS devices [3]. This technique enhances the phase noise while still utilizing an explicit MOS varactor. However, achieving a compromise trade-off between the phase noise, power consumption, and FTR is still a challenge. The inductive tuning method introduced in [7] also extends the tuning range. The inductor loaded transformer designed in [7] uses switches placed at different locations on the secondary coil to increase the number of sub-bands, and the resulting tuning range. However, the relatively large number of switches required for symmetric implementation increases the parasitic capacitance. As the number of switches increases, the total load capacitance (resistance) reflected back to the primary of the transformer increases (decreases). Furthermore, the parasitics of switches that are off still affects the performance. Due to these parasitics the structure has a relatively narrower frequency tuning range. In this work, we not only use less number of switches but also present a design approach which is based on the mutual coupling between the auxiliary coils. The proposed approach lowers the loading effect of the parasitics of switches when reflected back to the primary side, which in turn results in a wider tuning range and lower phase noise than those of [7]. In chapter 2, we will also show that by increasing the coupling factor among the loads the adverse effects of loading are reduced and thus lower loss and better phase noise can be achieved. The improved coupling is facilitated by proposing a transformer structure that uses three layers of metals to achieve a 3D transformer structure. 7 In this chapter, we propose a design technique for a mm-wave VCO architecture which incorporates a switchable transformer topology to extend the tuning range with minimal adverse effects on the phase noise performance and area. The switches in the secondary and tertiary coils are implemented differentially to reduce the on-state resistance that negatively affects the phase noise. Moreover, employing a three-dimensional (3D) implementation of the transformer, i.e., each coil is laid out in a different metal layer, the mutual coupling factors, especially one between the secondary and tertiary coils, are improved leading to lower the loading effect of the auxiliary sides on the primary coil. Given that NMOS Class-C VCOs typically have a better performance in terms of power consumption and phase noise compared to traditional cross-coupled LC-VCOs [11], we select this structure for the core oscillator. The proposed VCO offers a compact size, wide tuning range, and low phase noise. The rest of this chapter is organized as follows; Sections II and III describe the proposed technique and the VCO implementation, respectively. Measurement results are presented in Section IV and concluding remarks are provided in Section V. 2.2. Proposed Inductive Tuning Method [12] The proposed inductive tuning technique benefits from a switchable transformer topology that is comprised of three magnetically coupled inductors two of which are terminated by switches. In this method, by turning these switches on and off, the magnetic flux passing through the primary coil is varied, leading to a change in the effective inductance seen at the primary side. Because of the presence of the two switches, four frequency sub-bands are achieved as a result of turning the auxiliary coils on and off. In this setting, the fine tuning is realized by varactors to continuously cover the sub-bands introduced by the proposed inductive tuning method. To have a design guideline, a simplified model for the proposed inductive tuning method considering all the possible switching states is shown in Fig. 1. In this figure, kij, Rsi, Coff, and Rp 8 represent the coupling factor between the corresponding ith and jth coils, the loss (i.e., series resistance) of the ith inductor, the off-state parasitic capacitance of the switch, and the on-state resistance of the switch, respectively. To analyze the circuits shown in Fig. 1, without loss of generality, the following simplifying assumptions are considered; (i) the off-state resistance and on-state parasitic capacitance of the switches are negligible. (ii) The loss of the secondary and tertiary coils are assumed to be the same (𝑅𝑆2 = 𝑅𝑆3 = 𝑅𝑆), and (iii) the mutual inductance between the secondary and primary coils is the same as that of the tertiary and primary coils (𝑀12 = 𝑀13 = 𝑀). L1L2RS2RonK12Zin V1+-V2+-L3RS3RonV3+-K13 K23L1L2RS2CoffK12Zin V1+-V2+-L3RS3CoffV3+-K13 K23ZL2ZL3ZL3ZL2L1L2RS2RonK12Zin V1+-V2+-L3RS3V3+-K13 K23ZL2ZL3CoffL1L2RS2K12Zin V1+-V2+-L3RS3V3+-K13 K23ZL2ZL3CoffRon(a) (b)(c) (d) Figure 1. The simplified model for the proposed inductive tuning technique; (a) SW1 and SW2 are off; (b) SW1 and SW2 are on; (c) SW1 is on and SW2 is off; (d) SW1 is off and SW2 is on. The simple model of the proposed transformer is shown in Fig. 1. Note that this circuit model consider the high frequency loss which is summarized in RSi. As mentioned earlier, the objective of using the simplified model is to provide some insights and intuitions on the advantages of the proposed technique. The high-frequency effects (parasitics) will definitely affect the performance, however, including them when we just want to motivate the idea will complicate the derivations and perhaps will cause more confusion. 9 To validate the effectiveness of the proposed transformer, consider the simplified model of a non-ideal transformer shown in Fig. 2. L1L2MV2+-L3 V3+-MM23I1I3I2ZL3ZL2Zin V1+- Figure 2. Simplified schematic of the proposed transformer. According to Fig. 2, we have; 1 1 1 2 3V j L I j MI j MI (1) 2 1 2 2 23 3V j MI j L I j M I (2) 3 1 23 2 3 3V j MI j M I j L I (3) 2 2 2LV Z I (4) 3 3 3LV Z I (5) Then, from Eq. (2) to (5), we have; 2 2 233 2 1 23 3 23[ ]LLj L Z j MI I T Ij L Z j M (6) By substituting Eq. (6) in (1), 2 1 2 12 2 23 1[ ]Lj MI I T Ij L Z j M T (7) where T1 and T2 are defined as; 2 2 2313 3 23LLj L Z j MTj L Z j M (8) 10 22 2 23 1Lj MTj L Z j M T (9) By substituting Eq. (6) and (7) in (1), the input impedance of Zin can be found as; 1 2 1(1 )inZ j L j MT T (10) 2 22 3 2 3 231 2 2 223 2 3 2 3 2 3 3 2( 2 )L LinL L L LM j L j L Z Z j MZ j LM L L Z Z j L Z j L Z (11) Note that equation (10) (when expanded to (11)) is symmetric with respect to L2 and L3 and is valid for all switch states. However, the simplifying assumptions, which are valid and applicable in the frequency band of interest, mentioned next may affect the symmetry of the equations. In the case where both switches are on, ZL2 and ZL3 are equal to (RS+RON). This term is relatively small and is negligible as compared to other terms in both the numerator and denominator of Eq. (8), and thus T1 becomes, 2 2313 23L MTL M (12) Then, the input impedance in this case is calculated as, 2 21 12 23 1(1 )inLMZ j L Tj L Z j M T (13) Next, defining A and B as; 2 3 233 23[ 2 ]L L MAL M (14) 2 23233 23L MB ML M (15) The final equation of the input impedance can be written as; 2 212. .( ) ( )inS ONA MZ j LR R j L B (16) 11 Equation (16) is not symmetric with respect to L2 and L3 (it can be shown that it has a symmetric imaginary part while its real part is not symmetric), however, in the frequency band of interest (16) offers a good approximation of the original equation which is symmetric. In the case where both switches are off, ZL2 and ZL3 are equal to (𝑅𝑆 −𝑗𝜔𝐶𝑜𝑓𝑓). Given that the quality factors are high, RS can be neglected as compared to other terms. Thus, the simplified T1 becomes, 2 12 231 2 13 23( )( )offoffL C MTL C M (17) Then, the input impedance in this case can be calculated as; 2 21 112 23 1(1 )( )inoffMZ j L Tj L j C j M T (18) Again, By defining A’ and B’ as; 2 12 3 23'2 13 23[ 2( ) 2 ]( )offoffL L C MAL C M (19) 2 12 23'23 2 13 23( )( )offoffL C MB ML C M (20) In this case, the input impedance can be written as; ' 2 21 ' 12. .( ) ( )inoffA MZ j Lj L B j C (21) This equation is indeed symmetric with respect to L2 and L3. In the case where SW1 is on and SW2 is off, we have 𝑍𝐿2 = 𝑅𝑜𝑛 + 𝑅𝑆 and 𝑍𝐿3 =1𝑗𝜔𝐶𝑜𝑓𝑓+ 𝑅𝑆. Thus, the input impedance of Zin (please refer to Eq. (10)) can be found as; 12 2 21 12 23 1(1 )( )inon SMZ j L Tj L R R j M T (22) where T1 is defined as; 2 231 2 13 23( )offL MTL C M (23) Thus, the equivalent inductance and resistance can be found as; 2 122 3 231 2 2 123 2 23 3 232 13 232 ( ).( ) ( )( )offeqoffoffL L M CML LM L M L C MLL C M (24) 2 12 22 3 232 2 1223 2 23 3 232 13 232 ( ). .( )( ) ( )( )( )offeq S onoffoffL L M CMR R RM L M L C MLL C M (25) Equation (25) indicate that the coupling of the third coil still impacts the equivalent quality factor at the primary even when the switch is off. The additional terms containing M23 in (25) is manipulated to further lower Req in this sub-band. Similar behavior can be seen for the remaining sub-bands. The assumption of M12=M13=M was made as it greatly simplifies the expressions and also in the implemented prototype, the values of M12 and M13 are indeed close to each other. However, this is not a necessary design choice and in general M12 and M13 can be different. Eqs. (24) and (25) represent the case where the secondary coil is shorted and the tertiary one is open. For the other case where the secondary coil is open and the tertiary one is shorted, we have; 21L SoffZ Rj C (26) 3L S ONZ R R (27) Thus, the equivalent inductance and resistance can be found as; 13 2 122 3 231 2 2 123 3 23 2 233 12 232 ( ).( ) ( )( )offeqoffoffL L M CML LM L M L C MLL C M (28) 2 12 22 3 232 2 1223 3 23 2 233 12 232 ( ). .( )( ) ( )( )( )offeq S onoffoffL L M CMR R RM L M L C MLL C M (29) By comparing Eqs. (28) and (24) and considering that L2 is not equal to L3, we can conclude that the center frequency of the second and third sub-bands are different. It is worth noting that in our design L2 and L3 (and also k12 and k13) are not equal, however, M12 is approximately equal to M13. Refer to Eq. (10), the input impedance at the primary coil becomes, 2 212 2 23(1 )inLMZ j L Tj L Z j M T (30) where ω is the frequency and T defined as: 2 2 233 3 23LLj L Z j MTj L Z j M (31) Fig. 1(b) illustrates the case in which both switches are on. In this circuit, taking into account assumption (ii), we have 𝑍𝐿2 = 𝑍𝐿3 = 𝑍𝐿. Substituting 𝑍𝐿 by (𝑅𝑜𝑛 + 𝑅𝑆), the equivalent inductance and resistance seen by the primary coil are written as; 2 2 2 23 2 232 233 23 3 231 1 22 2 22 232 233 23.(1 ).( . )( ) ( . )eqS onL M L MM L ML M L ML L L CL MR R L ML M (32) 2 2 2 233 2312 2 22 232 233 23. .(1 ).( )( ).( ) ( )S oneq S onS onL MM R RL MR R R CL MR R L ML M (33) where C1 and C2 are defined as; 14 𝐶1 =𝑀2(1 +𝐿2 −𝑀23𝐿3 −𝑀23)(𝐿2 +𝑀23𝐿2 −𝑀23𝐿3 −𝑀23)2 (34) 𝐶2 =𝑀2(1 +𝐿2 −𝑀23𝐿3 −𝑀23)(𝐿2 +𝑀23𝐿2 −𝑀23𝐿3 −𝑀23) (35) The approximations in (32) and (33) hold if (𝑅𝑜𝑛 + 𝑅𝑆) is small compared to the second term in the denominator which is a reasonable assumption in the frequency band of interest. From (32) and (33), one can manipulate C1 and C2 to improve Req and in turn the resulted quality factor at the cost of a slightly degraded tuning range. Shown in Fig. 3, C1 and C2 are plotted versus the coupling factor between the secondary and tertiary coils (k23) for given design parameters. This figure implies that an increase in the value of k23 drops the equivalent resistance while slightly increasing the minimum achievable inductance, thereby improving the quality factor. For instance, if k23 increases from 0.3 to 0.5, the equivalent resistance reduces by ~20% while the corresponding inductance increases by only ~7%. Figure 3. Plotted C1 and C2 versus k23 with L2=210 pH, L3=250 pH, and M=115 pH. 15 The intuitive verification of the above concept can be explained using Fig. 4 (a) and (b). This figure presents two ideal cases where 𝑘23 = 0 and 𝑘23 = 1. In Fig. 4(a), assuming 𝐿2 = 𝐿3 = 𝐿 and 𝑍𝐿2 = 𝑍𝐿3 ≈ 𝑅𝑜𝑛, the equivalent inductance and resistance seen by the primary coil are derived as 𝐿𝑒𝑞 = 0.5𝐿 − 𝑀 and 𝑅𝑒𝑞 = 0.5𝑅𝑜𝑛 while the corresponding values for the same parameters are obtained as 𝐿𝑒𝑞 = 𝐿 −𝑀 and 𝑅𝑒𝑞 = 0.5𝑅𝑜𝑛 in Fig. 4(b) where 𝑘23 = 1. This clearly shows that the quality factor seen by the primary side is doubled approximately if the two circuits are perfectly coupled. This confirms our previous observation that increasing k23 results in improving the quality factor of the tank. L1-M 0.5L-M0.5RonM(a)L1-M L-M0.5RonM(b)Zin Zin Figure 4. Simplified π-model of the transformer when; (a) K23=0. (b) K23=1. Fig. 5 shows the schematic diagram of the cases where both switches are either on or off (that is, ZL2=ZL3). For simplicity, we assume that L2=L3=L and RS2=RS3=RS. Thus, we have V2=V3 and I2=I3=I. Also, we assume that the transformer is designed such that the coupling factors are approximately the same, that is, M12M13M, L1L2MV2+-L3 V3+-MM23I1I3I2ZL3ZL2Zin V1+- Figure 5. Simplified model of the transformer where both switches are on (or off). we have, 16 1 1 1 2 3V j L I j MI j MI (36) 2 1 2 23 3V j MI j LI j M I (37) 3 1 23 2 3V j MI j M I j LI (38) If we assume the following equivalent circuit (Fig. 6) for three-coil structure: L1-M Leq-MReqMV1+-V2+-i2i1 Figure 6. Simplified pi-model for one of transformers. According to Fig. 6, we then have 1 1 1 2V j L i j Mi (39) 2 1 2eqV j Mi j L i (40) If k23=0, from Eq. (36) to (38) we have; 1 1 1 (2 )V j L I j M I (41) 2 1 (0.5 )(2 )V j MI j L I (42) By comparing Eq. (41) with (39), and also (42) with (40), the right hand side of (42) should be equal to that of (40). Then; Leq=0.5L and i2=2I=2I2. Also, we have 𝑉2𝐼2=𝑉2𝐼= 𝑅𝑂𝑁 and 𝑅𝑒𝑞 =𝑉2𝑖2 . Considering that i2=2I=2I2, we can conclude that Req=0.5RON. For the other case where k23=1, 1 1 1 (2 )V j L I j M I (43) 2 1 (2 )V j MI j L I (44) 17 By comparing Eq. (44) with (40), and also (43) with (39), the right hand side of (44) should be equal to that of (40). Then; Leq=0.5L and i2=2I=2I2. Also, we have 𝑉2𝐼2=𝑉2𝐼= 𝑅𝑂𝑁 and 𝑅𝑒𝑞 =𝑉2𝑖2 . Considering that i2=2I=2I2, we can conclude that Req=0.5RON. As mentioned earlier and also according to the aforementioned analysis, increasing the coupling factor between the loads can improve the phase noise by reducing the loading effect of the switches. To increase k23, the second and third coils are implemented in different adjacent metal layers. This is to circumvent the minimum distance between two adjacent traces that are in the same metal layer (as dictated by the design rules). Furthermore, such an implementation can satisfy M13M12=M as the secondary and tertiary coils are separated from the primary coil by an approximately equal distance. Note that k23 is not linked to the primary coil. In other words, to increase K23, instead of utilizing the same M8 layer to implement the two auxiliary coils, we implement them in M8 and AP. In this configuration, the minimum distance between two adjacent traces with the same metal layer, enforced by DRC, is irrelevant as the coils are on different metal layers. Therefore, the coupling factor is improved due to the smaller distance between the implemented coils. Thus, the proposed three dimensional transformer achieves a higher coupling factor as compared to the simple planar implementation. Additionally, such an implementation can satisfy M13=M12=M as the secondary and tertiary coils are separated from the primary coil by an approximately equal distance (in the technology used in this work). The proposed transformer is built using the topmost three metal layers. In such a configuration, as compared to planar implementation, the coupling factor is improved and the occupied area is reduced. From (32) and (33), the equivalent inductance and resistance at the primary coil are also dependent on the values of L2 and L3 in addition to k23. Figs. 7 and 8 show the plot of C1 and C2 for different values of L2 and L3. As can be seen from the plots, to reduce the loading effect of the switches, L3 18 should increase. In other words, for a given k23, increasing L3 lowers Req. However, increasing L3 to any arbitrary high value is not reasonable as it leads to a larger occupied area, higher parasitic capacitance, and higher loss with only marginal improvement in Req beyond a certain point. Therefore, L3 is optimized to achieve a low phase noise while having minimal impact on the tuning range. Fig. 9 shows that k remains almost intact if the original structure of the coils is preserved even if L3 varies. L3 Figure 7. Simulated C1 versus different values of L2 and L3. L3 Figure 8. Simulated C2 versus different values of L2 and L3. 19 L36055 Figure 9. Dependency of the coupling factor on L3. 2.3. Summary of The Design Procedure The two additional coils that are added to the main (primary coil) and can be switched on or off results in 4 distinct sub-bands (off-off, on-off, off-on, and on-on states). The 4 possible states are analyzed by presenting the equivalent inductance and resistance seen by the primary coil in each state. As a summary of design procedure, after designing a conventional oscillator, the continuous frequency tuning range (FTR) in each sub-band should be determined by the varactor size (taking into account the trade-off between the TR and Q of the varactor). To implement the secondary coil, the first and third sub-bands should be considered. L2 is chosen to satisfy the target continuous FTR. Then, L3 should be designed to cover the gap between the 1st and 3rd sub-bands. The next step is to maximize the coupling factor between the coils especially k23. The next step is to iterate (if necessary) the design parameters to meet the design specifications and optimize the performance taking into account the trade-off between the PN and TR. Based on the formulas discussed earlier, the value of L2 and L3 are optimized based on the width and the choice of the 20 metal layer for each coil. Figs. 7 and 8 can help the designer to gain an insight about the optimum value of L2 and L3. The value of the inductances are chosen (based on the formulas) to achieve a wide tuning range with minimal adverse effect on the phase noise. Finally, design and layout techniques are used to achieve the maximum quality factor over the desired frequency band. The quality factor and effective inductance simulation results are presented in Figs. 10 and 11, respectively. Figure 10. The simulated Q-factor over all switch states. Figure 11. Simulated effective inductance over the whole frequency tuning range. 21 As can be seen, when both switches are off, the auxiliary coils have minimum effect on the primary coil due to the relatively high-impedance of off-state parasitic capacitance of switches, and thus, the quality factor and effective inductance in the off-off state are the highest amongst the all possible states. When SW1 is off and SW2 is on, the outermost coil, which is implemented by AP, is shorted while L2 is open. This configuration achieves the second highest effective inductance and quality factor due to the higher quality factor of AP and also lower coupling factor due to larger distance to the primary coil. When SW1 is on and SW2 is off, because of the higher coupling factor which is the corollary of a shorter distance to the primary coil, the effective inductance of this sub-band is lower than that of the previous sub-band. However, the quality factor is the worst amongst that of other states due to the lower quality factor of M8 and also lower effective inductance of this sub-band. Finally, for the state when both switches are on, the effect of both auxiliary coils are added to further reduce the effective inductance. The quality factor of this setting is better than that of the previous sub-band (SW1 on and SW2 off), because the effective inductance is reduced while the loading effect is further reduced (we can intuitively say that the loads are in parallel which in turn results in a lower loss), which reduces the Q-factor degradation. Thus, the on-on state achieves a lower effective inductance along with a larger quality factor as compared to on-off state. Moreover, according to Fig. 10, the Q-factor differences among the four states would impact the performance. However, in our design, we are trying to reduce this effect by proposing a three dimensional tank. The 3rd and 4th sub-bands (on-off state and on-on state, respectively) have the worst case quality factor (refer to Fig. 10) due to larger loading effect and also lower effective inductance (refer to Fig. 11). The PN performance is further degraded in such sub-bands as the oscillation frequency is higher. To ameliorate these issues, we have implemented the auxiliary coils in different metal layers to increase the coupling factor (k23) and in turn improve the quality 22 factor in the 2nd, 3rd ,4th sub-bands. Besides, the AP layer can achieve a higher quality factor compared to M8 thanks to its thicker metal and larger skin depth. Thus, the third coil is implemented in AP layer rather than M8 to improve the PN in the two sub-bands (2nd and 4th) to which the third coil contributes. 2.4. VCO Design And Implementation The schematic of the proposed VCO is shown in Fig. 12 (a). In this figure, M1 and M2 are the cross-coupled transistors in the gm-cell of the VCO. Cd1 and Cd2 are chosen large enough to lower the loop gain degradation. The combination of control voltage and digital inputs allows for having fine and coarse control over the range of oscillation frequency. To achieve a better quality factor, and in turn, phase noise, a larger size switch should be utilized [13-14]. However, at mm-wave frequencies, the off-state parasitic capacitance of the switch limits the tuning range. To ameliorate this limitation, NMOS switches are designed differentially to lower the corresponding on-state resistance, thereby achieving a better phase noise. The switch sizes are optimized to achieve a reasonable trade-off between phase noise and tuning range. The schematic of the implemented switch is presented in Fig. 12(b). 23 CCd1M1 M2VDDL1CVARCVARVCTRLCd2VOUT+VOUT-RBIASVBIAS VBIASL2SW1L3SW2K13K12K23M1 M2EnableRBIAS180°/0° 0°/180° M3(a)(b) Figure 12. Schematic of the proposed (a) VCO. (b) The proposed switch architecture. To achieve a lower loss and thus a better Q, the thicker topmost metal layers are employed to implement mm-wave inductors. In the 65-nm CMOS process used in this work, the thickest metal layer is metal 9 (M9). Although implementation of mutually coupled planar inductors utilizing M9 shows a higher quality factor, there are challenges associated with such a planar configuration. First, the occupied area is larger due to the minimum required spacing between side-by-side M9 layers enforced by the design rules. This also results in a lower coupling factor, and in turn, a higher magnetic loss. Second, a relatively large number of required vias to avoid metal overlaps introduces higher losses. Third, the current crowding (i.e., the proximity effect) of the adjacent inductor traces results in a lower quality factor at mm-wave frequencies. Finally, there is a parasitic capacitance between adjacent traces of planar inductors which limits the self-resonance frequency (SRF) and quality factor at high frequencies. The proposed mutually coupled inductors ameliorate these challenges by implementing the three coupled inductors in three top-most metal layers, that is, Metal 8 (M8), M9, and the top aluminum layer (AP). The M8 layer is utilized to implement the secondary coil where a larger inductance with smaller inner radius is required due to have a higher inductance per unit length. Moreover, the skin effect degrades the quality factor at high 24 frequencies. In this regard, AP has the advantage of a larger skin depth and thus a lower quality factor degradation. Furthermore, the substrate loss is reduced if AP layer is employed (due to its farther distance from the substrate). The cross-sectional and top views of the proposed switchable transformer which consists of three mutually coupled one-turn inductors are shown in Figs. 13 and 14, respectively. The design parameters r, Wi, and Si are the inner radius, width of ith inductor, and the spacing between adjacent traces, respectively. The inner radius and widths of the inductors are chosen to satisfy the conditions for the inductances mentioned in the previous section. S2 and S3 are chosen to be zero in order to maximize the coupling factor without increasing the inter-layer parasitic capacitance. S1 is chosen to maximize the quality factor of the overall tank at the target frequency. d3d1d2S2S3S12rW3W2W1 W1M9AP M8 Substarte Figure 13. Cross section view of the proposed switchable transformer. 25 SW1SW2VDDGm-cellW3 W2 W1 W1 S1S2S32RM9AP M8 Figure 14. Top view of the proposed switchable transformer. The center-tapped inductor is designed in the tank to facilitate DC biasing and save the area. Moreover, to lower the parasitic resistance, a striped inductor structure is used to improve the phase noise [16]. In the proposed transformer, a grounded shield ring using M9 is designed to further enhance the quality factor at the desired frequency. This shielding layer significantly reduces the loss associated with the lossy substrate at high frequencies which results in a higher quality factor. This structure functions similar to a coplanar waveguide in which the space between the ground and signal planes is optimized to achieve the maximum possible quality factor at the frequency band of interest. The optimized distance in our design is 25 μm. This structure functions similar to a coplanar waveguide (CPW) in which the space between the ground and signal planes is optimized to achieve the maximum possible quality factor at the frequency band of interest. Electromagnetic (EM) simulations are performed using Keysight’s Advanced Design System (ADS) Momentum to validate the performance of the proposed transformer architecture. Figs. 15 and 16 show the simulated coupling factor, k, and Q for various possible configurations, respectively. In these configurations, the secondary coil is implemented in M8 layer and the tertiary 26 coil is either implemented with M8 (planar structure) or with AP (the proposed structure). For a fair comparison, the spacing between the secondary and tertiary coils is chosen to be the minimum distance allowed by the design rules for the planar implementation. Note that in the proposed design we have used S2=S3=0, which would further improve the coupling factor. Figs. 10 and 11 show the simulated effective inductance and quality factor over the entire 4 sub-bands. Figure 15. Simulated coupling factor (k23) for different configurations. Figure 16. Simulated quality factor for different configurations. The layout of the fabricated VCO is completely symmetric. This feature is facilitated by using the following design techniques. The differential switches shown in Fig. 12(b) are used to keep the 27 symmetry of the layout and to reduce the on-state switch resistance. The main switch is located exactly in the middle to keep the symmetry of the layout. The switches connected to ground and the Gm-cell are mirrored to keep the vertical symmetry of the layout. The primary coil is implemented with its center-tap connected to the supply voltage. The secondary and tertiary coils are designed using standard symmetric inductors provided by the technology. 2.5. Verification Of The Proposed Model Although the proposed circuit model of the proposed three coils transformer is simple, however, as it has been mentioned above, the objective of using the simplified model is to provide some insights and intuitions on the advantages of the proposed technique. The high-frequency effects will definitely affect the performance, however, including them when we just want to motivate the idea will complicate the derivations and perhaps will cause more confusion. In other words, the simplified model is used to simplify analysis and the guidance for the simulations. A more elaborate model of an individual inductor is shown in Fig. 17. This model is more appropriate for modelling the high-frequency effects. To analyze an inductor individually, the high-frequency effects should certainly be considered. However, when analyzing a circuit including inductors, such as a VCO, especially when the frequency of operation is (much) lower than the self-resonance frequency, considering only L and R as a model of inductor is a reasonable choice, because in such cases one can ignore the effect of parasitic capacitance CX , which affects the self-resonance frequency (SRF). Moreover, the substrate loss and parasitic capacitance CP affect the quality factor of the inductor which is considered in the simulation results shown in Fig. 18. 28 CXL RCPCPRSub RSub12CSub CSub Figure 17. A high frequency circuit model of an inductor. To further justify the statements mentioned earlier, we present simulation results for two cases of inductors modelled in Keyshight’s ADS momentum on the layout of the proposed transformer and the case using the simplified R and L model for each coil of the transformer with constant mutual coupling in the schematic simulations. The schematic view simulation results of the simplified model (in which L and R of each inductor are extracted from EM simulation of the corresponding individual inductor) are compared to EM simulation results of the laid out transformer. Fig. 18 compares the simulated quality factor seen by the primary coil in two cases when k23 = 0 and 0.5, respectively. As can be seen from the figure, the quality factor seen by the primary coil is improved as the coupling factor between the loads increases (in this case from 0 to 0.5). Both EM simulations on the layout and simulations on the simplified schematic view confirm the same trend. Thus, although the quality factor extracted from the simplified model is different from that of the EM layout (because it does not take into account high-frequency effects), however, the trend predicted in equations derived from the simplified circuit are still valid and the quality factor improves as the coupling factor increases. Same conclusion can be drawn for the effective inductance seen at the primary. 29 Figure 18. Q factor comparison between that of the proposed model and transformer. 2.6. Discussion On The Pulling Effect Of The Proposed VCO In this design, by turning switches on and off, the output loading of the VCO suddenly changes, which could result in the change of frequency of operation. Thus in what follows we will address the issue of the frequency pulling. In this work, there are four switching states and all of them have been modeled. In the proposed model, the off-state and on-state effects of the switches are considered and then the oscillation frequency in each state is analyzed by taking into account the inductance and resistance that are seen by the tank in each state. The analysis shows that the oscillation frequency in all sub-bands are controllable by the user utilizing two bits. To validate the transient response, a pulse signal with a rising edge (which turns on the switch) is applied to one of the switches at 15 ns and the simulated output voltage and oscillation frequency are shown in Figs. 19 and 20, respectively. These plots show that the VCO settles to its steady state after ~0.3 ns. 30 Figure 19. Simulated output signal after applying a pulse at the gate of the switch. Figure 20. Transient response of the oscillation frequency after applying a pulse at the gate of the switch. 31 Chapter 3. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO 3.1. Introduction and Overview of Recently Published Works Scaling of the cost effective CMOS process attracts designers to implement mm-wave building blocks in recent years. Due to the insatiable demand for high data-rate communication capacity, the radio-frequency integrated circuits (RFICs) operating in higher frequency bands has attracted more attention. In this regard, the 5th generation of wireless systems (5G) has extended its operation frequency to millimeter-wave (mm-wave) band in order to benefit from more signal bandwidth [1]. Voltage-controlled oscillators (VCOs) play an important role in the operation of integrated mm-wave transceivers. The generation of a wideband, low phase noise signal while maintaining low power consumption is one of the challenges in designing RF and mm-wave VCOs. Furthermore, power consumption is one of the most important challenges in designing mm-wave VCOs. As the frequency goes up, the noise performance of active and passive devices deteriorates the phase noise of the VCO. Also, due to the trade-off between power consumption and phase noise, more power should be consumed to have an acceptable noise performance. Several techniques have been introduced to improve the tuning range and phase noise performance of different VCO topologies [21–25]. To extend the tuning range in [21], even and odd operation modes of a transformer-based resonator are combined to generate two overlapping frequency sub-bands. However, in this technique achieving an overlap between the two bands becomes a major challenge at high frequencies. Moreover, due to the complicated fabrication processes and layout, utilizing integrated passive device (IPD) process, used in [21], is challenging for applications with operation frequencies higher than 10 GHz. An inductor-loaded transformer is designed in [22] with switches placed at different positions to increase the number of sub-bands, and the resulting tuning range. However, the relatively large number of switches required for a 32 symmetric implementation increases the parasitic capacitance seen by the resonator. In [8], switchable coupled VCO cores are designed to realize a dual mode operation at the cost of a higher power consumption and larger area. The design also suffers from a low output power. In [9][15][26], switched coupled inductors are utilized to change the effective inductance of the tank, thereby extending the tuning range at the cost of a degraded noise performance. The design also requires a relatively large area and high power consumption. In [27], by using analog tuning varactors connected to the gm-cell a wide tuning range is achieved, however, such varactors have a poor performance at mm-wave frequencies. In this thesis, we introduce a new structure which simultaneously improves the tuning range and phase noise while consuming lower power as compared to the conventional RF and mm-wave Colpitts VCOs. The tuning range is enhanced by using a switched-capacitor bank which has less adverse effects on the noise performance. In the proposed architecture, the switches in the capacitor bank minimally load the tank. That is, during the switching action, low impedance path to ground which would degrade the tank quality factor is avoided. In addition, two inter-stage inductors between the tank and Gm-cell are employed to relax the start-up issues at the cost of a larger area. The proposed approach not only reduces the power consumption, but also improves the tuning range by partially cancelling the parasitic capacitances seen by the tank. Also, the dynamic body bias technique is applied to the VCO core transistors to improve the phase noise and to lower the power consumption. Given that Colpitts VCOs show better behavior in terms of noise performance than those of traditional cross-coupled LC-VCOs [7], we have used the Colpitts structure for the core oscillator. To confirm the effectiveness of the proposed techniques, the small-signal analysis in conjunction with simulation results in different regimes of operation are presented. Furthermore, the 33 performance of the overall architecture is validated through measurements of a proof-of-concept prototype chip fabricated in a 65-nm CMOS process. The organization of the rest of the paper is as follows: Sections II and III describe the proposed techniques and the VCO implementation, respectively. Measurement results are presented in Section IV and concluding remarks are provided in Section V. 3.2. The Proposed Architecture 3.2.1. The Proposed Variable Capacitive Feedback Colpitts [23] The quality factor of the tank has a pronounced effect on the phase noise of the oscillator. Typically, the tank quality factor is limited by the quality factor of the tank inductor at low frequencies and by that of the varactor at mm-wave frequencies. Thus, at high frequencies, the varactor should be chosen as small as possible in order to minimize the degradation of the tank quality factor. To achieve a wide-tuning range and overcome this limitation, discrete tuning method by using digitally controlled capacitor banks have been used [13-14]. However, the design of switches and capacitors has its own challenges (such as the effect of parasitic capacitances of switches at high frequencies as well as loading the tank). To ameliorate these challenges, several techniques such as differential capacitor banks and differential switches have been proposed [13-14]. Although these techniques have shown better performance with respect to the single-ended switching stage, they load the tank by creating a low impedance path to ground. In addition, the conventional capacitor banks typically use large switches (to minimize the switch resistance) which in turn adversely affect the tuning range. An advantage of the proposed switching structure is that as compared to the conventional capacitor banks, the parasitic capacitance of the switches seen by the tank is lower. Furthermore, in the proposed technique in which the capacitor bank is being added in parallel with one of the feedback capacitors, the effective incremental capacitance 34 of the capacitor bank can be made lower. This would result in a lower rate of frequency change (i.e., lower VCO gain, kVCO) which in turn would lead to a better frequency resolution and a lower sensitivity to noise on the control voltage [33]. Moreover, due to this lower rate of frequency change, the proposed structure has a better suppression of the flicker noise [4][16][17]. To be more specific, for the same value of the branch capacitance in the proposed and conventional capacitor banks, the coarse tuning range due to activating each branch of the capacitor bank would be narrower in the proposed technique, thereby requiring a smaller varactor with lower KVCO for fine tuning (to cover the overlap between different frequency bands). To summarize, the proposed technique offers lower flicker noise, higher frequency resolution, and lower sensitivity to the noise of the control voltage. The schematics of both conventional and the proposed capacitor banks, implemented in a differential Colpitts VCO, are shown in Fig. 21 (a) and (b). CVDDLCVARCVARLVCTRLC1C2C1C2M1 M2IBIAS IBIASC3 2C3CVDDLCVARCVARLVCTRLC1C2 C2M1 M2IBIAS IBIASC1 C4(a)(b)C32C3C4C4C4 Figure 21. Schematic of the Colpitts VCO with (a) Conventional (b) Proposed capacitor bank. To have a fair comparison of the phase noise between these two VCOs, we consider the same frequency of oscillation and switch size for both cases. In the conventional structure, because of 35 the extra capacitor that is in parallel with the output node, the oscillation frequency will be lower than that of the proposed circuit, if they both have the same incremental capacitor values. Thus, one can choose a bigger capacitor value for the proposed technique to have the same oscillation frequency. To analyze the circuits shown in Fig. 22 (a) and (b), the following simplifying assumptions and circuit parameters are considered: 1 2C C c (45) 3 4ON ON ONR R R (46) 3 4C CQ Q Q (47) 2 2 232 2 231 ONONP ONONC RR RC R (48) 33 2 2 231PONCCC R (49) 1ONPRc (50) 1 2 1 4 231 2 1 2 4( )PPPC C C C CCC C C C C (51) 36 LPM1 C1C2IBIASRPVOUTVXC3RON3LPM1C1C2IBIASRPC4RON4VXVOUTCeq LP RP CP3 RONP RNEGGNDC1C2RONPCP4VOUTGNDVOUT(c)(d)-VOUT -VOUT(a) (b)LP RP CP_NEW RONP_NEW RNEGGNDVOUTVXYT Figure 22. Half-circuit and small-signal models of : (a, c) conventional switching network (b, d) proposed switching network. It is worth mentioning that as shown in (47) we assume that the quality factor of the capacitors of the bank in both approaches (refer to Fig. 22 (a) and (b)) are the same. As given by Equation (50), we assume that the impedance of the Colpitts feedback capacitor, C1, is much lower than the impedance of the switches and thus, the switches do not impact the start-up condition. Note that this condition is usually satisfied at high frequency of operation. Equation (51) is based on the assumption that the oscillation frequency is equal in both conventional and the proposed structures. In addition, RONP and CP3 are the equivalent parallel resistance and capacitance of the series RC network in the conventional circuit. The small-signal model of conventional and proposed Colpitts oscillators with a one-branch capacitor bank are illustrated in Fig. 22 (c) and (d), respectively. The total admittance of the network illustrated in Fig. 22 (d) can be expressed as in (52). The real and imaginary parts indicate the total parallel resistance and capacitance seen by the tank as stated in (53) to (55). By comparing (48) and (54) which show the equivalent parallel resistance of the two structures, it can be concluded that in the conventional capacitor bank the quality factor of the tank is worse than that of the proposed approach due to having a lower equivalent parallel 37 resistance. The equivalent parallel resistance in the proposed technique is at least 3 times larger than that of the conventional bank. Therefore, this technique can achieve lower phase noise. Another advantage of the proposed approach is that for the same perceived parallel resistance (which translates to approximately the same phase noise) the proposed approach can use a smaller switch. Furthermore, because of cascaded switches in the proposed approach (refer to Fig. 21), the effect of the parasitic capacitance of the switches at the output node is significantly decreased. This, in turn, would result in a wider tuning range. 2 22 242 34 42 2 241(2 )( )(2 )1 (2 )TONP PONPONP P PP ONPcYR c CRc R c c C c Cjc C R (52) 2 22 24Re[ ]1(2 )TONP PONPcYR c CR (53) _ 2 214ONP NEW ONPONPR RR c 24 424ONP P ONP PR C R Cc c (54) 2 24 4_ 2 2 24( )(2 )1 (2 )ONP P PP NEWP ONPc R c c C c CCc C R (55) Based on the above-mentioned analysis, the proposed capacitor bank can be added in parallel to either C1 or C2. However, due to the following three reasons, it makes more sense to have the bank in parallel with C1. First, in order to suppress the flicker noise up-conversion, the bias current of the VCO and the parasitic capacitance at the sources of transistors M1 and M2 should be as small 38 as possible [13-14]. Adding the capacitor bank to C1 will result in less parasitic capacitance at the source of transistors M1 and M2. Second, the parasitic capacitance at the sources of M1 and M2 is larger than that of the drains of these transistors, thus adding the capacitor bank to C2 will result in further increase of the parasitic capacitance and thus reduces the tuning range. Third, based on (58), increasing C1 reduces the noise factor and in turn the phase noise, while increasing C2 would increase these parameters. Therefore, increasing C1 would be a better choice for the purpose of this work where the focus is on lowering the phase noise. Note that in (56) to (58), ‘A’,’ ’, and ‘F’ are the oscillation amplitude, power efficiency, and noise factor of the conventional Colpitts oscillator, respectively. 21 22 BIAS PCA I RC C (56) 21 2C AC C VDD (57) 211CFC (58) In other words, the conventional capacitor banks typically load the tank by creating a low impedance path to ground which significantly degrades the noise performance. To overcome this challenge, one can use large switches to minimize the on-state resistance of switches which, in turn, adversely affect the tuning range by increasing the switch off-state capacitance. To ameliorate these challenges, the capacitor bank based on variable capacitive feedback technique, i.e, the capacitor bank is being added in parallel with one of the feedback capacitors, is introduced in [23]. As shown in [23] by using the variable capacitive feedback technique, larger value of perceived parallel resistance, i.e., less loading, can be achieved compared to a conventional topology using 39 similar switch sizes in the tank. This technique improves the tuning range with a minimal degradation of the tank quality factor. The schematics of both conventional and the proposed capacitor banks, implemented in a differential Colpitts VCO, for simplicity only one capacitor branch is shown in Figs. 23 (a) and (b). CVDDLCVARCVARVCTRLC1C2C1M1 M2CM3 M4(a)C3 C3CVDDLCVARCVARVCTRLC1C2C1M1 M2CM3 M4(b)C4C4 Figure 23. Schematic of the conventional differential Colpitts VCO with a one-branch capacitor bank (a) conventional; (b) proposed. As discussed earlier, the equivalent parallel resistance seen by the tank in the conventional and proposed structures can be summarized here as; 2 2 23_ 2 2 231 ONONP conv ONONC RR RC R (59) _ _ 2 2_14ONP proposed ONP convONP convR RR c 2_ 4 _ 424ONP conv P ONP conv PR C R Cc c (60) where ω, RON, RONP. CP4, and c are the frequency, on-state resistance of the switch, the equivalent parallel on-state resistance of the switch, the equivalent parallel capacitance of the added capacitor in the proposed topology, and the value of feedback capacitors (𝑐 = 𝐶1 = 𝐶2), respectively. Comparing (59) and (60) reveals that the equivalent parallel resistance achieved by the tank in the proposed approach is at least three times larger than that of the conventional one due to the additional terms of (60). To show the effectiveness of this switching technique, a combination of 40 post-layout and electromagnetic (EM) simulations using Cadence and Keysight Advanced Design System (ADS) Momentum are performed. The simulated loss of the tank (Rp) is plotted in Fig. 24 for the conventional and proposed topologies and over different switch states. This plot suggests that the proposed method has a lower tank loss when the switch is on which in turn results in a lower phase noise. Figure 24.The simulated loss of the tank in conventional and proposed structures. 3.2.2. The Proposed Gm-boosting Technique The conventional Colpitts VCOs usually have a high power consumption and also suffer from start-up issue [28][31-32][34-37]. To overcome these challenges, the negative transconductance generated by the Gm-cell should be improved. A typical solution to boost the negative transconductance is to increase the injection current to the tank by increasing the size of the cross-coupled transistors. However, larger power would be required to generate larger injection current which would result in a degraded figure-of-merit (FOM) suggested for VCOs [25][27][30]. The use of large devices also limits the tuning range of VCO due to their large parasitic capacitances. 41 An alternative method is to decrease the capacitive loading of the tank in order to achieve a larger transconductance. To this end, we propose a Gm-boosting technique that splits the tank inductor (as shown in Fig. 25) to achieve an improved negative transconductance with a low power consumption and a wide tuning range. CVDDLCVARCVARVCTRLC1C2C1M1 M2CM3 M4(a)CVDDL1CVARCVARVCTRLC1C2C1M1 M2CM3 M4(b)L2 L2CBCBCB CB Figure 25. Schematic of the (a) conventional Colpitts VCO; (b) proposed gm-boosted Colpitts VCO. The proposed Gm-boosting technique significantly enhances the generated negative transconductance by adding two inter-stage inductors between the Gm-cell and the tank. In other words, the splitting of the tank inductor into two parts and placing the varactors between them is utilized in the tank in order to relax the start-up of the VCO while reducing the parasitic capacitance and power consumption. As can be seen in Fig. 25, the inductor L2 is used to distribute the capacitive loading of the tank. The output buffers are connected to the interconnection node between L1 and L2 and their parasitic capacitance has been modeled by CB. To analyze the effect of the inter-stage inductors, simplified schematic diagrams and small-signal models of the conventional and proposed structure are shown in Fig. 26 (a, b, c, and d). In these figures Cp, Geq, 42 and Ceq are the drain parasitic capacitance of M1, equivalent parallel conductance of the Gm-cell, and the equivalent parallel capacitance of the Gm-cell, respectively. C1 M1M32C2-1-1 CpL2CBYin_proposed(b)CB Geq Ceq(c)CpYin_Conv CB Geq Ceq(d)CpYin_proposedL2C1 M1M32C2-1-1 CpCBYin_Conv(a) Figure 26. The simplified and small signal model of the (a, c) conventional Colpitts VCO; (b, d) proposed Colpitts VCO. According to Fig. 26(c), the real and imaginary parts of the input admittance of the conventional VCO can be written as; _Re[ ]in Conv eqY G (61) _Im[ ] ( )in Conv B p eqY C C C (62) Similar calculations can be carried out for the circuit of Fig. 26(d) and the corresponding parameters are as follows: 43 2 22_ 2 22 2 221Re[ ]1[ ( ) ]eqin proposedeqeq p eqGLYG LG C CL (63) _Im[ ]in proposed BY C 22 22 221( )( )[ ]1[ ( ) ]p eqp eqeq p eqC CLC CL G L C CL 22 2 22 2211[ ( ) ]eqBeq p eqGCLL G L C CL (64) The approximations in (63) and (64) hold if 𝜔(𝐶𝑝 + 𝐶𝑒𝑞) −1𝜔𝐿2≈ 0 which means that L2 should resonate with (Cp + Ceq) within the frequency band of interest. Comparing (61) and (63), we advocate that the equivalent negative transconductance in the proposed approach can be designed to be much larger than that of the conventional one within the frequency band of interest. Also, it can be shown that the value of L2 required to satisfy the above-mentioned condition also leads to 1𝐺𝑒𝑞𝜔2𝐿22 > 𝐺𝑒𝑞 which means that the proposed approach offers a larger equivalent conductance. Moreover, from (62) and (64), the proposed method also reduces the overall parasitic capacitance seen by the tank. As can be observed from the first term of expression in (64), for all values of L2 the coefficient of 𝜔(𝐶𝑝 + 𝐶𝑒𝑞) is less than one (and also the third term is negative) showing that the imaginary part of admittance is smaller as compared to (62). The simulated small-signal equivalent negative transconductance versus frequency is plotted in Fig. 27. As can be seen, the 44 generated negative transconductance is boosted significantly at the frequency band of interest provided by tuning the value of L2. Figure 27. Comparison of the simulated small-signal negative transconductance in conventional and proposed topology. Due to the fact that oscillators show a non-linear large signal behavior, the large signal S-parameters simulation (LSSP) is performed to validate the aforementioned analysis. Fig. 28 and 29 present the simulated large-signal negative resistance and transconductance of the proposed gm-cell versus the applied differential voltage at 28-GHz. These plots show that the negative resistance decreases when the amplitude increases such that the resistance becomes positive beyond a certain point. This also would result in a sharp transition in the transconductance showing that the oscillation amplitude must be lower than the zero crossing point in order to satisfy start-up conditions. 45 Figure 28. Comparison of simulated equivalent parallel resistance in Conventional and proposed technique. Figure 29. Comparison of simulated equivalent parallel transconductance in Conventional and proposed technique. Another interesting advantage of the proposed technique is that the inter-stage inductors act as an impedance transformer network as shown in Fig. 30 (a). As discussed earlier, the total admittance seen by the tank is substantially larger than that seen by L2, which translates to 𝑎𝑏𝑠(𝑌1) > 𝑎𝑏𝑠(𝑌2). 46 Applying ohm’s law and considering that 𝐼1 = 𝐼2 = 𝐼 (assuming that Cp is compensated mostly by L2 and does not draw current), it can be concluded that 𝑉2 > 𝑉1. This observation asserts that the swing voltage across the gate of M1 is also boosted resulting in a sharper edges and faster switching of the cross-coupled pair hence a lower phase noise. The simulated transient signals at both ends of L2 are presented in Fig. 30 (b). To summarize, contrary to the conventional Gm-boosting techniques, the proposed technique simultaneously enhances the negative transconductance, lowers the power consumption and phase noise, and extends the tuning range at the cost of a larger occupied area. TankVDDL2 L2gm-cellI1=Y1*V1I2=Y2*V2Y1Y2(a) (b) Figure 30. System-level model of the proposed gm-boosting technique. (b) Simulated transient signals at V1 and V2. 3.2.3. The Proposed Body Bias Technique The proposed dynamic forward-body self-biased technique is shown in Fig. 31 (a). In this approach, the bodies of the transistors, including M1, M2, M3, and M4, are biased through a cross-coupled structure. 47 CVDDLCVARCVARVCTRLC1C2C1M1 M2CM3 M4C1M1M3(a)(b) Figure 31. The proposed body bias technique; (a) schematic (b) corresponding signals. There are several advantages associated with the proposed topology compared to the conventional one, i.e. all substrate nodes connected to ground. First, the forward body bias reduces the threshold voltage of the transistors which results in a larger generated negative transconductance. The simulated large signal negative transconductance generated at 28-GHz in conventional and proposed topology are plotted and compared in Fig. 32. Second, this technique also improves the noise performance of the VCO. To better illustrate this, the corresponding waveforms are depicted in Fig. 31 (b) and Fig. 33. 48 Figure 32. Comparison of the simulated negative transconductance in conventional and proposed body bias technique. Figure 33. Simulated transient response of the proposed body bias technique. As can be seen from Fig. 31 (b), the gate voltages of M3 and M4 are in phase with the corresponding signals applied to their bodies and their sources are connected to the ground. Thus, the threshold voltage variation only depends on the variation of the body signal. Moreover, the threshold voltage variation is in opposite phase with that of the body signal. Therefore, when the gate voltage 49 decreases, the threshold voltage increases to turn the corresponding transistor off faster. On the other way, when the gate voltage rises, the threshold voltage goes down to turn the transistor on faster. Fig. 33 shows the transient threshold and gate-source voltages of M3 verifying the above-mentioned observations. This feature facilitates a sharper transition between on and off-state of the cross-coupled pair reducing the conduction time of the transistors, and in turn leads to a more impulse-shaped drain current compared to that of the conventional topology. This not only improves the phase noise, but also lowers the power consumption. Fig. 34 shows the simulated transient drain current of M1. Figure 34. The simulated transient drain current of M1. In this plot, given some certain values, the current has a negative value which is the effect of the parasitic capacitances. The frequency of oscillation in the proposed technique is slightly lower due to existing more parasitic capacitance associated with the bulk. The average current per cycle in the proposed technique is lower than that of the conventional which results in a lower power consumption. Finally, the first and third harmonic of the output current are in opposite phases shown by its behavior at its positive peak values which is one of the characteristics of the square waves. Overall, the proposed body bias technique facilitates sharper switching which 50 simultaneously lowers conduction time and, in turn, power consumption which leads to a better noise performance. 3.2.4. VCO Design and Implementation The proposed VCO architecture is shown in Fig. 35. The transistors M1 and M2 form the main cross-coupled pair of the gm-cell of the proposed VCO generating a negative resistance in order to compensate for the loss of the tank. The capacitors C1 and C2 realize the capacitive feedback network of the Colpitts oscillator which are selected to be equal so as to have a compromise trade-off between the noise performance and output power [13]. The cross-coupled pair of M3 and M4 form current switches which result in an improved generated negative resistance by providing larger loop gain [8], although they limit the oscillation amplitude. Moreover, adding M3 and M4 in the source of the main cross-coupled pair lowers the 1/f3 phase noise [13][14]. The NMOS switches SW1 and SW2 are employed to effectively change the feedback coefficient of the Colpitts capacitive feedback network. Depending on the switch states, three overlapping frequency sub-bands are introduced to extend the tuning range. Additionally, by avoiding the use of any switches connected between the output node and ground, the degradation due to lossy switches are considerably lower. The combination of control voltage and digital inputs allows for having fine and coarse control over the range of oscillation frequency. Considering process variations, due to the trade-off between the tuning range and quality factor of varactors, length and width of MOS varactors are tuned just large enough to guarantee non-zero overlap between the adjacent frequency sub-bands in all process corners. Two open-drain output buffers are designed at the tank output for measurement purposes. The drain biases are provided using external Bias-Tees through a separate 1-V supply voltage. To avoid limiting the tuning range by increasing the parasitic 51 capacitance, the buffers are connected directly to the VCO output without employing DC-coupling capacitors. CVDDL1C1C2C1M1 M2CM3 M4CaddedCadded CaddedCaddedL2 L2Enable 1 Enable 2CVARCVARVCTRLEnable 1Enable 2VDDLBTCBT50 Ω M6External Bias-TeeVDDLBTCBT50 Ω M6External Bias-Tee Figure 35. Schematic of the fabricated VCO. To sustain oscillation at high frequencies without increasing the transistors’ size, the inter-stage inductors, L2, are designed. These inductors boost the negative transconductance and extend the tuning range, resulting in a low-power wide-tuning-range VCO. The center-tapped inductor is designed in the tank to facilitate DC biasing and save area. Moreover, to reduce the parasitic resistance and, in turn, achieve a higher quality factor, a striped inductor structure [12] is used in the tank. The top and cross sectional view of the implemented coil are shown in Figs. 36 (a) and (b), respectively. 52 2r RRRRRRRRVDDPOS NEGW W SSGNDGNDGNDGND(a) (b)dS2rW WM9 SubstarteRShieldR Figure 36. Different views of the proposed tank; (a) top; b) cross sectional. The design parameters are r, W, S, and R which are the inner radius, width of the two parallel inductor traces, space between the two traces, and the outer radius, respectively. They are optimized to achieve the maximum possible quality factor in the frequency band of interest. Furthermore, a grounded shield ring using M9 is designed to further enhance the quality factor at the desired frequency. This shielding layer significantly reduces the loss associated with the lossy substrate at high frequencies which results in a higher quality factor at the cost of lower inductance and higher parasitic capacitance. Full-wave Electromagnetic simulations using ADS Momentum are performed to evaluate the performance of the proposed tank. 53 Considering the process variations, the performance of the proposed VCO is simulated in different process corners. In order to improve the accuracy of simulations, a combination of post-layout and electromagnetic simulations are performed using Cadence and ADS Momentum. The simulation results are summarized in Table I verifying that the proposed VCO shows a robust performance over all process corners. Table 1. The simulated performance summary of the proposed VCO over the process corners. Corner SS TT FF TR (GHz) 24.8-29.4 25.1-29.9 25.3-30.3 TR (%) 17 17.5 18 Average PN (dBc/Hz) –118.4 –120.3 –121.5 Average Power (mW) 1.65 2.3 3.35 Average FOM (dBc/Hz) –183.9 –184.4 –184.9 Average FOMT (dBc/Hz) –188.5 –189.2 –190 54 Chapter 4. Measurement Results 4.1. Low-Phase-Noise Wide-Tuning-Range Mm-wave CMOS VCO To validate the proposed architecture, a VCO with a tuning range from 50.1 to 59.8 GHz is designed and implemented in a 65-nm CMOS process. Fig. 37 shows the micrograph of the fabricated VCO. The area of the circuit excluding the pads is 260 μm × 250 μm. The supply voltage is 1 V and the power consumption for the VCO core is 6.2 mW. Two open-drain output buffers, biased using external Bias-Tees, are designed at the tank output for measurement purposes. These buffers collectively draw 8 mA from a 1-V supply. The VCO output is measured using on-wafer probing while the DC biasing of the circuit is supplied by bond wires. The measured output spectrum is illustrated in Fig. 38 showing that the center frequency of the fabricated VCO is ~55 GHz with the calibrated output power of –6.4 dBm. At the center frequency, VCO achieves a phase noise of –89.9 and –111.72 dBc/Hz at 1 and 10-MHz offset, respectively as depicted in Fig. 39. The measured PN at 10-MHz offset over the entire frequency tuning range is plotted in Fig. 40. The measured calibrated output power over the entire frequency tuning range is also shown in Fig. 41. Furthermore, based on Fig. 42, over different states of the switches, the tuning range including four frequency bands is over 18% resulting in a figure of merit (FOM) and FOM with tuning range (FOMT) of –178.9 and –184 dBc/Hz at 10-MHz offset, respectively. Table III summarizes the performance of the proposed design along with comparison to that of the state-of-the-art designs also implemented in 65-nm CMOS. As can be seen from the table, the performance of the proposed design, particularly in terms of FOM and FOMT compares favorably with other designs. 55 GG VDDSGGGGSVB VTEN1 EN2260um250um Figure 37. Micrograph of the fabricated 60-GHz VCO. -6.38@54.71 GHz–10–20–30–40–50 Figure 38. Output spectrum of the fabricated VCO. The fabricated VCO is measured using a Rohde & Schwarz (R&S) FSW 26.5 GHz spectrum analyzer. A V-band external harmonic mixer, R&S Z75, is used as the frequency extender. The loss of the external mixer, probe (ZPROBE GSG 67 GHz), and the cable are measured and then inputted into the spectrum analyzer to de-embed the measured data. As mentioned in the paper, 56 two open-drain output buffers are designed at the VCO output for the measurement purpose which draw a total of 8 mA current from a 1-V supply. The phase noise reported here is measured at the center frequency of oscillation. In our design, this reported phase noise is close to the average phase noise over the tuning range. The measured phase noise at 10-MHz offset over the entire frequency tuning range is also plotted in Fig. 40. Based on this figure, the best, average, and the worst case phase noise are –114.2, –111.9, and –109.9 dBc/Hz, respectively. -111.72@10 MHz-89.90@1 MHz–100 dBc/Hz–102 dBc/Hz–104 dBc/Hz–106 dBc/Hz–108 dBc/Hz–110 dBc/Hz–112 dBc/HzFreq = 54.917183740 GHz Figure 39. Measured phase noise at the center frequency of oscillation. 57 Figure 40. The measured PN at 10-MHz offset over the whole frequency tuning range. Figure 41. The measured output power over the whole frequency tuning range. 58 Figure 42. Measured frequency tuning range over different control voltages. 4.2. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO To validate the proposed architecture, a 28-GHz VCO suitable for 5G applications is designed and implemented in a 65-nm CMOS process. Fig. 43 shows the micrograph of the fabricated VCO laid in a CFP-24 package with bonding wires. The area of the circuit excluding the pads is 460 μm × 480 μm. The supply voltage is 1 V and the power consumption for the VCO core is 2.3 mW. The output buffers collectively draw 10 mA from a separate 1-V supply. The measurements were performed with the package mounted on a printed circuit board (PCB) and tested using on-wafer probing along with the bond wires for the DC biasing of the circuit. Also, large values of decoupling capacitors is realized using SMD capacitors placed in parallel between the supply voltage (VDD) and ground to relax the VCO from the supply noise in different frequency bands. The measured output spectrum is illustrated in Fig. 44 showing that the center frequency of the fabricated VCO is ~26 GHz with an uncalibrated output power of –17.8 dBm. The total loss of the probe and cable is measured as ~3 dB in the frequency band of interest. At the center frequency, 59 VCO achieves a phase noise of –86 and –122.1 dBc/Hz at 300-KHz and 10-MHz offset, respectively as depicted in Fig. 45. Furthermore, based on Fig. 46, over different states of the switches and sweeping the control voltage, the tuning range including three frequency bands is over 20%, from 23.7-GHz to 29-GHz resulting in a FOM and FOM with tuning range (FOMT) of –187 and –193 dBc/Hz at 10-MHz offset from the center frequency of oscillation, respectively. Table IV summarizes the performance of the proposed design along with comparison to those of the state-of-the-art designs. As can be seen from the table, the performance of the proposed design, particularly in terms of phase noise, power, FOM, FOMT, and FOM with considering the chip area (FOMA) compares favorably with state-of-the-art. 460 μm480 μmGGSGGSVCtrl GEN1 EN2G VDD GVCtrl Figure 43. Micrograph of the fabricated 28-GHz VCO. 60 Figure 44. Measured output spectrum of the fabricated VCO. Figure 45. Measured phase noise of the fabricated VCO at the center frequency of oscillation. 61 Figure 46. Measured tuning range of the fabricated VCO. To evaluate the robustness of the fabricated VCO, the measured phase noise variation over the entire tuning range is plotted in Fig. 47. This figure shows that phase noise has a small variation of 2.8 dBc/Hz over the frequency band of interest. 2.8 dBc/Hz Figure 47. Measured PN and FOM at 10-MHz offset from the carrier over the entire tuning range. 62 To validate the sensitivity of the proposed VCO performance on the supply voltage, the measurement results for different values of supply voltage is presented in Table II. According to this table, although the overall performance is better for 1-V supply, the VCO still shows a satisfying performance for a supply of 0.8-V verifying the usefulness of the proposed VCO in both low-voltage and low-power applications. Note that for supplies less than 0.8 V, the VCO would not oscillate. Table 2. The performance summary of the fabricated VCO for different supply voltages. Supply (V) 0.8 0.9 1 TR (GHz) 25.5-30.2 25.1-30 23.7-29 TR (%) 16.9 17.8 20.1 Average PN (dBc/Hz) –116.6 –119.1 –121 PN Variation (dBc/Hz) 9.6 5.3 2.8 Average Power (mW) 1.25 1.8 2.3 Average FOM (dBc/Hz) –184.8 –185.1 –185.8 Average FOMT (dBc/Hz) –189.3 –190.1 –191.8 63 Chapter 5. Conclusions 5.1. Low-Phase-Noise Wide-Tuning-Range Mm-wave CMOS VCO A transformer-based mm-wave VCO with wide tuning range and low phase noise is presented. It is shown that optimization of the secondary and tertiary coils inductances and the coupling factor between them improve the phase noise substantially as compared to the conventional designs. By switching in and/or out the secondary and the tertiary coils of the transformer, 4 overlapping frequency sub-bands are implemented which collectively result in a wide tuning range. Design techniques that are proposed to implement a high-Q tunable inductor for the tank allow the VCO to achieve a competitive phase noise performance. As a proof of concept, a 55-GHz VCO is implemented in a 65-nm CMOS process. The following Table summarizes the performance of the proposed design and compares it with that of the state-of-the-art. Table 3. Performance summary of 60-GHz VCOs in 65-nm CMOS. VCO [7] [18] [19]* [20] This work Center Freq. (GHz) 61 77 63 66.1 55 Average PN @10 MHz (dBc/Hz) –108.3 –108.4 –92.2 –103 –111.9 TR (%) 14.2 15.8 24.6 27.9 18 Power (mW) 6 14.3 21.5# 13 6.2 1FOM@10MHz (dBc/Hz) –176.2 –174.5 –174.8 –168.3 –178.9 2FOMT@10MHz (dBc/Hz) –179.3 –178.5 –182.6 –177.2 –184 *Measured at 1-MHz offset from the carrier frequency #Including the output buffer 1 | 20log[ ] 10log[ ]1offsetosc DCfoffsetf PFOM PNf mW 2 (%)20log[ ]10TTRFOM FOM 64 5.2. Low-Power Wide-Tuning-Range Mm-wave CMOS VCO A mm-wave VCO architecture based on Colpitts design and achieving wide tuning range and low phase noise is presented. A novel capacitor bank design substantially minimizes the phase noise degradation as compared to the conventional designs. By switching in and/or out the capacitor bank, 3 overlapping frequency sub-bands are implemented which collectively result in a wide tuning range. Moreover, inter-stage inductors are designed and placed between the gm-cell and the tank as the gm-booster. The proposed technique not only relaxes the start-up and power consumption issues common to conventional Colpitts oscillators, but it also improves the tuning range by reducing the parasitic capacitance seen by the tank. Furthermore, a body bias technique is utilized to further reduce the power consumption and improve the phase noise. As a proof of concept, a 28-GHz VCO suitable for 5G applications is implemented in a 65-nm CMOS process. Design techniques proposed to implement a high-performance mm-wave VCO allows for achieving a competitive FOM, FOMT, and FOMA in comparison with state-of-the-art designs. Table IV summarizes the performance of the fabricated VCO. Table 4. Performance comparison with that of the state-of-the-art designs. VCO Tech (nm) Freq (GHz) PN (dBc/Hz) Power (mW) TR (%) Area mm2 FOM1 (dBc/Hz) FOMT2 (dBc/Hz FOMA3 MTT’17 [1] 65 26.3 -119.1 4.3 40.3 0.25 -180 -192.1 -186 MWCL’14 [9] 65 27.5 –113 6 43.3 0.81 -174.3 –187 -175.2 TCASII’13[25] 130 40 –95 12 27 1.93 -176.3 –184.9 -173.4 MTT’17 [26]* 90 20.85 -100.7 8.1 15.8 0.48 -177.8 -181.8 -183.3 IMS’09 [28]* 130 23 -100 4 2 0.32 -181 -167 -186 MWCL’17 [29]* 180 24 -101.3 25# 12.6 0.23 -174.9 -176.9 -181.5 This work 65 26.3 –122.1 2.3 20.1 0.22 –187 –193 -193.6 *Measured at 1-MHz offset from the carrier frequency #Including the output buffer 1 | 20log[ ] 10log[ ]1offsetosc DCfoffsetf PFOM PNf mW 2 (%)20log[ ]10TTRFOM FOM 232( )10log[ ]1AArea mmFOM FOMmm 65 Future Work The performance of the proposed VCOs can be further improved in terms of efficiency and output power. For efficiency, one can investigate techniques to improve the DC to RF efficiency of the proposed mm-wave VCO. In the context of output power, one can explore design techniques for providing output power in the range of 0 dBm and more to eliminate the use of LO buffer stages. Eliminating the buffers in turn help reducing the overall power consumption. Another interesting path is to look into techniques to increase the operation frequency. The operation frequency extension to 100 GHz and beyond requires elaborate techniques as the operation frequency becomes closer to the cut-off frequency of the transistors. For example, in the technology used in this research, which is 65-nm CMOS technology, the cut-off frequency is about 120 GHz. Designing a robust VCO operating in frequencies in the vicinity or higher than 120 GHz is still challenging. 66 References [1] P. Agarwal et al., “Switched Substrate-Shield-Based Low-Loss CMOS Inductors for Wide Tuning Range VCOs,” IEEE Transactions on Microwave Theory and Techniques, vol. 65, no. 8, pp. 2964-2976, Aug. 2017. [2] Z. Zong, M. Babaie, and R. B. Staszewski, “A 60 GHz 25% tuning range frequency generator with implicit divider based on third harmonic extraction with 182 dBc/Hz FoM,”in Digest of Papers - IEEE Radio Frequency Integrated Circuits Symposium, 2015, vol.2015–Novem, pp. 279–282. [3] L. Li, P. Reynaert, and M. Steyaert, “A 60-GHz CMOS VCO using capacitance-splitting and gate–drain impedance-balancing techniques,” IEEE Trans., 2011. [4] F. Pepe, A. Bonfanti, S. Levantino, C. Samori, and A. L. Lacaita, “Analysis and minimization of flicker noise up-conversion in voltage-biased oscillators,” IEEE Trans. Microw. Theory Tech., vol. 61, no. 6, pp. 2382–2394, Jun. 2013. [5] H. Shimomura, “A Study on High-Frequency Performance in MOSFETs Scaling,” 2011. [6] J. H. Lee, et al., “60 GHz CMOS downconversion mixer with 15.46 dB gain and 64.7 dB LO-RF isolation,” Electronics Letters, vol. 49, no. 4, pp. 264-266, Feb. 14 2013. [7] W. Fei, H. Yu, H. Fu, J. Ren and K. S. Yeo, “Design and Analysis of Wide Frequency-Tuning-Range CMOS 60 GHz VCO by Switching Inductor Loaded Transformer,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 61, no. 3, pp. 699-711, March 2014. [8] Q. Zou, K. Ma and K. S. Yeo, “A Low Phase Noise and Wide Tuning Range Millimeter-Wave VCO Using Switchable Coupled VCO-Cores,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 62, no. 2, pp. 554-563, Feb. 2015. [9] J. Zhang, N. Sharma and K. K. O, “21.5-to-33.4 GHz Voltage-Controlled Oscillator Using NMOS Switched Inductors in CMOS,” IEEE Microwave and Wireless Components Letters, vol. 24, no. 7, pp. 478-480, July 2014. [10] H. K. Chen, H. J. Chen, D. C. Chang, Y. Z. Juang, and S. S. Lu, “A 0.6 V, 4.32 mW, 68 GHz low phase-noise VCO with intrinsic-tuned technique in 0.13 μm CMOS,” IEEE Microw. Wirel. Components Lett., vol. 18, no. 7, pp. 467–469, 2008. [11] Hajimiri and T. H. Lee, “A general theory of phase noise in electrical oscillators,” IEEE Journal of Solid-State Circuits, vol. 33, no. 2, pp. 179-194, Feb 1998. 67 [12] M. H. Kashani, A. Tarkeshdouz, R. Molavi, E. Afshari, and S. Mirabbasi, "A Wide-Tuning-Range Low-Phase-Noise mm-Wave CMOS VCO with Switchable Transformer-based Tank," IEEE Solid-State Circuits Letters, vol. 1, no. 3, pp. 66-69, June 2018. [13] Razavi. RF microelectronics.Vol. 1, New Jersey, Prentice Hall, 1998. [14] H. Darabi, Radio Frequency Integrated Circuits and Systems, Cambridge University Press, 2015. [15] M. Demirkan, S. P. Bruss and R. R. Spencer, “Design of Wide Tuning-Range CMOS VCOs Using Switched Coupled-Inductors,” IEEE Journal of Solid-State Circuits, vol. 43, no. 5, pp. 1156-1163, May 2008. [16] H. Tsuji, et al, “A study of flicker noise suppression of K-band VCO using striped inductor,” Asia-Pacific Microwave Conference (APMC), Nanjing, pp. 1-3, Nov. 2015. [17] Q. Shi, D. Guermandi, J. Craninckx and P. Wambacq, “Flicker noise upconversion mechanisms in K-band CMOS VCOs,” 2015 IEEE Asian Solid-State Circuits Conference (A-SSCC), pp. 1-4, Xiamen, 2015. [18] G. Liu, R. Berenguer and Y. Xu, “A MM-Wave Configurable VCO Using MCPW-Based Tunable Inductor in 65-nm CMOS,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 58, no. 12, pp. 842-846, Dec. 2011. [19] N. Yanay and E. Socher, “Wide Tuning-Range mm-Wave Voltage-Controlled Oscillator Employing an Artificial Magnetic Transmission Line,” IEEE Transactions on Microwave Theory and Techniques, vol. 63, no. 4, pp. 1342-1352, April 2015. [20] C. H. Hung and R. Gharpurey, “A 57-to-75 GHz dual-mode wide-band reconfigurable oscillator in 65nm CMOS,” IEEE Radio Frequency Integrated Circuits Symposium, Tampa, FL, pp. 261-264, June 2014. [21] S. Wang and P. H. Chen, “A Low-Phase-Noise and Wide-Tuning-Range CMOS/IPD Transformer-Based VCO With High FOMT of −206.8 dBc/Hz,” in IEEE Transactions on Components, Packaging and Manufacturing Technology, vol. 6, no. 1, pp. 145-152, Jan. 2016. [22] W. Fei, H. Yu, H. Fu, J. Ren and K. S. Yeo, “Design and Analysis of Wide Frequency-Tuning-Range CMOS 60 GHz VCO by Switching Inductor Loaded Transformer,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 61, no. 3, pp. 699-711, March 2014. 68 [23] Milad Haghi Kashani, et al. “ A Wide-Tuning-Range Low-Phase-Noise Colpitts Oscillator with Variable capacitive Feedback” IEEE International Symposium on Circuits and Systems (ISCAS), pp. 1-4, Florence, Italy, May 2018. [24] S. Hu, F. Wang and H. Wang, “A transformer-based inverted complementary cross-coupled VCO with a 193.3dBc/Hz FoM and 13kHz 1/f3 noise corner,” IEEE Radio Frequency Integrated Circuits Symposium (RFIC), pp. 27-30, San Francisco, CA, June 2016. [25] Q. Wu et al., “Frequency Tuning Range Extension in LC-VCOs Using Negative-Capacitance Circuits,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 60, no. 4, pp. 182-186, April 2013. [26] Z. Chen et al., “Linear CMOS LC-VCO Based on Triple-Coupled Inductors and Its Application to 40-GHz Phase-Locked Loop,” IEEE Transactions on Microwave Theory and Techniques, vol. 65, no. 8, pp. 2977-2989, Aug. 2017. [27] T. P. Wang and S. Y. Wang, “Frequency-Tuning Negative-Conductance Boosted Structure and Applications for Low-Voltage Low-Power Wide-Tuning-Range VCO,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 23, no. 6, pp. 1137-1144, June 2015. [28] C. K. Hsieh, K. Y. Kao, J. R. Tseng and K. Y. Lin, “A K-band CMOS low power modified colpitts VCO using transformer feedback,” IEEE MTT-S International Microwave Symposium Digest, pp. 1293-1296, Boston, MA, 2009. [29] Y. T. Chang and H. C. Lu, “A K-Band High-Efficiency VCO Using Current Reused Technique,” IEEE Microwave and Wireless Components Letters, vol. 27, no. 12, pp. 1134-1136, Dec. 2017 [30] T. P. Wang and S. Y. Wang, “Frequency-Tuning Negative-Conductance Boosted Structure and Applications for Low-Voltage Low-Power Wide-Tuning-Range VCO,” IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 23, no. 6, pp. 1137-1144, June 2015. [31] K. W. Ha, H. Ryu, J. H. Lee, J. G. Kim and D. Baek, “Gm -Boosted Complementary Current-Reuse Colpitts VCO With Low Power and Low Phase Noise,” IEEE Microwave and Wireless Components Letters, vol. 24, no. 6, pp. 418-420, June 2014. 69 [32] J. P. Hong and S. G. Lee, “Low Phase Noise Gm-Boosted Differential Gate-to-Source Feedback Colpitts CMOS VCO,” IEEE Journal of Solid-State Circuits, vol. 44, no. 11, pp. 3079-3091, Nov. 2009. [33] J. Kim, J. Shin, S. Kim and H. Shin, “A Wide-Band CMOS LC VCO With Linearized Coarse Tuning Characteristics,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 55, no. 5, pp. 399-403, May 2008. [34] N. Mahalingam, K. Ma, K. S. Yeo and W. M. Lim, “K-band High-PAE Wide-Tuning-Range VCO Using Triple-Coupled LC Tanks,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 60, no. 11, pp. 736-740, Nov. 2013. [35] S. L. Liu, K. H. Chen, T. Chang and A. Chin, “A Low-Power K-Band CMOS VCO With Four-Coil Transformer Feedback,” in IEEE Microwave and Wireless Components Letters, vol. 20, no. 8, pp. 459-461, Aug. 2010. [36] Yin Fei Meng, B. M. Frank and A. El-Gabaly, “A novel variable inductor-based differential Colpitts VCO design with 17% frequency tuning range for 30 and 60 GHz applications,” 2014 IEEE MTT-S International Microwave Symposium (IMS2014), Tampa, FL, 2014, pp.1-4. [37] T. P. Wang, “A K-Band Low-Power Colpitts VCO With Voltage-to-Current Positive-Feedback Network in 0.18 μm CMOS,” IEEE Microwave and Wireless Components Letters, vol. 21, no. 4, pp. 218-220, April 2011.
- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- On the design of high-performance mm-wave CMOS VCOs
Open Collections
UBC Theses and Dissertations
Featured Collection
UBC Theses and Dissertations
On the design of high-performance mm-wave CMOS VCOs Haghi Kashani, Milad 2018
pdf
Page Metadata
Item Metadata
Title | On the design of high-performance mm-wave CMOS VCOs |
Creator |
Haghi Kashani, Milad |
Publisher | University of British Columbia |
Date Issued | 2018 |
Description | The increasing demand for higher data rates in wireless communication such as IEEE 802.15.3c wireless personal area networks (WPANs), wireless local area networks (WLANs), and next generation wireless systems, have made the development of radio-frequency (RF) and millimeter-wave (mm-wave) frequency integrated circuits (ICs) that operate in few tens of gigahertz more popular. In the context of high data rate wireless systems, the 5th generation (5G) of wireless systems is expanding its operation frequency to millimeter wave (mm-wave) bands where more signal bandwidth are available. To implement such circuits, the preferred technology is CMOS (complementary metal-oxide semiconductor) since compared to other technologies, it offers lower supply voltage, lower cost, higher levels of integration, and potentially lower power. Many CMOS solutions for RF and mm-wave applications have been proposed over the past few years and thanks to continuous reductions in feature size of CMOS devices into deep-submicron range, many efficient and high performance state-of-the-art RF and mm-wave receivers have been reported. However, efficient system integration at mm-wave frequencies in CMOS is still a challenging task. Voltage-controlled oscillators (VCOs) are indispensable in the operations of fully integrated transceiver architectures. In many cases, quadrature local oscillator (LO) signals are required for frequency conversion. Major challenges in particular in portable applications include the generation of such quadrature phase outputs with low power consumption. This also applies to emerging mm-wave applications, where one would like to achieve a wide tuning range while maintaining low-power consumption and low phase noise. In this work, we investigate several design techniques for achieving a high performance wireless receiver building block with a specific focus on LC-based VCOs. Based on the measurement results, one of the VCOs achieves an average PN of −111.9 dBc/Hz at 10 MHz offset over the entire frequency tuning range, and a TR of ~18%, from 50.1 to 59.8 GHz, resulting in a figure of merit incorporating the tuning range (FOMT) of –184 dBc/Hz. The VCO core consumes 6.2 mW from a 1-V supply and excluding the pads occupies a compact silicon area of 0.06 mm2. |
Genre |
Thesis/Dissertation |
Type |
Text |
Language | eng |
Date Available | 2019-09-01 |
Provider | Vancouver : University of British Columbia Library |
Rights | Attribution-NonCommercial-NoDerivatives 4.0 International |
DOI | 10.14288/1.0371145 |
URI | http://hdl.handle.net/2429/66871 |
Degree |
Master of Applied Science - MASc |
Program |
Electrical and Computer Engineering |
Affiliation |
Applied Science, Faculty of Electrical and Computer Engineering, Department of |
Degree Grantor | University of British Columbia |
GraduationDate | 2018-09 |
Campus |
UBCV |
Scholarly Level | Graduate |
Rights URI | http://creativecommons.org/licenses/by-nc-nd/4.0/ |
AggregatedSourceRepository | DSpace |
Download
- Media
- 24-ubc_2018_september_haghikashani_milad.pdf [ 4.36MB ]
- Metadata
- JSON: 24-1.0371145.json
- JSON-LD: 24-1.0371145-ld.json
- RDF/XML (Pretty): 24-1.0371145-rdf.xml
- RDF/JSON: 24-1.0371145-rdf.json
- Turtle: 24-1.0371145-turtle.txt
- N-Triples: 24-1.0371145-rdf-ntriples.txt
- Original Record: 24-1.0371145-source.json
- Full Text
- 24-1.0371145-fulltext.txt
- Citation
- 24-1.0371145.ris
Full Text
Cite
Citation Scheme:
Usage Statistics
Share
Embed
Customize your widget with the following options, then copy and paste the code below into the HTML
of your page to embed this item in your website.
<div id="ubcOpenCollectionsWidgetDisplay">
<script id="ubcOpenCollectionsWidget"
src="{[{embed.src}]}"
data-item="{[{embed.item}]}"
data-collection="{[{embed.collection}]}"
data-metadata="{[{embed.showMetadata}]}"
data-width="{[{embed.width}]}"
async >
</script>
</div>
Our image viewer uses the IIIF 2.0 standard.
To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0371145/manifest