Sizing of multi-stage Op Amps by combining design equations with the gm/ID method☆
Introduction
Device sizing (including biasing) is an integral part of CMOS analog integrated circuit (IC) design. Due to the high dimensionality of the sizing parameters involved in many sizing problems, the designer often has to spend a great deal of time on circuit sizing.
Continuous down-scaling of the device feature size has driven the researchers’ attention to the design of multi-stage operational amplifiers (Op Amps) or operational transconductance amplifiers (OTAs) that can offer higher gains without impairing the signal swing, see for example [[1], [2], [3], [4], [5], [6]]. While higher gains can be achieved by multi-stage designs, design on their frequency compensations arises as an imperative design challenge.
In practice, experienced circuit designers would normally devote effects on establishing design equations to develop design insights and guide sizing. As a matter of fact, circuit-level design equations derived from alternative current (ac) analysis are the key means for developing the frequency compensation strategies in multi-stage Op Amp design. Some authors proposed to apply graphical means to guide the frequency compensation design [7,8], but it still ends up with qualitative design equations to guide sizing. Evidently, multiple-loop frequency compensation induces additional inter-stage constraints on the pole-zero placement via design equations. Such design equations must also be incorporated in sizing to ensure frequency performance. Without such design equations, multiple blind SPICE iterations would have to be incurred.
Recently, several symbolic methods have been proposed for automatically generating design equations in the ac domain [9,10]. These methods are particularly suited to the design of multi-stage amplifiers in that the effort of manual derivation of the pole-zero design equations can be substantially reduced. It is important to note that most design equations, although approximate, are analytical characterization of the dominant interrelations among a portion of design parameters. Such analytical relations can effectively reduce the blindness in searching feasible sizing regions. Unfortunately, most data-driven sizing optimization methods proposed in the literature such as [[11], [12], [13]] did not make use of analytical design equations in their optimization routines. Data-driven performance modeling or causality inference requires simulation-expensive routines.
Design equations can not only offer design insights, but also effectively reduce the uncertainty in sizing. However, due to the large number of design parameters involved in many sizing problems, using the circuit-level design equations only cannot completely determine all device sizes. Deterministic mapping from numerically derived device operating conditions to the device dimensions (W and L) with high reliability must be considered in initial sizing.
A traditionally used sizing method is to apply the following current-voltage (I–V) equation (using an n-channel MOSFET as an example) [14].where Cox is the gate oxide capacitance per unit area, μn is the mobility of n-type charge carriers, VTHn is the threshold voltage between the gate and source terminals, and W and L denote the channel width and effective channel length which are the sizing parameters of all MOSFETs throughout this paper. The aspect ratio always refers to W/L. Variables ID, VGS, and VDS carry their usual meanings. For devices in saturation two equations derivable from the above I–V equation areand(VDSAT denoting the saturation drain-source voltage), which are frequently used in the traditional sizing procedures, such as [[15], [16], [17]]. However, as pointed out by Razavi [14], these equations are mainly for developing intuition in analog design, but do not bear rigor or accuracy for sizing. Accurate initial sizing still has to resort to rigorous device model equations implemented in most SPICE simulators.
Several earlier publications on multi-stage Op Amp design outlined manual sizing procedures [[15], [16], [17]], which used the circuit-level design equations for noise, power, signal range, and frequency response performance as well as the device-level equations (2), (3) for calculating the MOSFET dimensions. Due to the inaccuracy of the device equations, many SPICE iterations are required to correct the calculated sizes and biasing, whereas the numbers of the correction times were often not reported in those publications.
An alternative to the device size mapping from the known operating conditions is by adopting a numerical table lookup tactic that builds a numerical device model profile by sweeping a set of selective device parameters. The gm/ID method is one such method that has received wide attentions. It can be used as a substitute to the simplistic square-law equations with the hope of reducing SPICE iterations in initial sizing.
The gm/ID method was published quite early [18,19], but did not receive wide spreading attention until quite recently more sophisticated design cases have been addressed [[20], [21], [22]]. Although these works mainly dealt with the single-stage amplifier designs, the robustness of the gm/ID method in generating accurate sizing results has been demonstrated. As one may observe from the work by Aminzadeh [22], more robust sizing results can be obtained if more sophisticated gm/ID sweeping data including variational sweeping data are incorporated. However, so far application of the gm/ID method to the sizing of multi-stage Op Amps is still rare. In Chapter 6 of Jespers and Murmann's book [23] the sizing of a two-stage fully differential Op Amp with Miller capacitor and nulling resistor compensation was developed by the gm/ID method. The sizing method developed there was targeted at applications to the switched-capacitor data converter circuits. Therefore, design specifications on the slew and settling metrics were more emphasized, for which the gain-bandwidth product (GBW) metric became the main design target. The authors introduced several MATLAB routines for running two-dimensional sweeping-based optimizations on several selected sizing parameters. However, since the proposed sizing method was centered on a specific circuit structure, it did not guarantee that the outlined highly customized sizing procedure could be easily extended to other multi-stage amplifier configurations.
The gm/ID sizing method promoted in this paper is targeted at initial sizing, which has its importance in several aspects. In practice, an initial sizing result can offer the designer with a good insight on what part of a circuit would need more attention for optimization. On the other hand, if several attempts of initial sizing could not give rise to a satisfactory circuit performance, it would be a good indication that probably the current circuit topology possesses certain defect. We illustrate the role of initial sizing in Fig. 1, showing that an efficient initial sizing method can not only benefit a subsequent sizing optimizer but also improve the circuit topology design in a topological synthesis loop.
Regarding sizing optimization, numerous previous works have made dedicated researches [13,[24], [25], [26], [27], [28], [29], [30], [31], [32]], with the majority of them requiring a sheer number of simulations. Fairly speaking, promising initial sizing results can substantially reduce the number of SPICE iterations in the subsequent optimization phase.
A systematic initial sizing procedure is developed in this paper for the class of multi-stage Op Amps (or OTAs) with the following features:
- 1.
We apply the gm/ID method with a variety of design equations to determine the initial sizing of a multi-stage Op Amp in a stage by stage fashion.
- 2.
We assume that the parasitic device capacitances are insignificant comparing to the compensation and load capacitors. Hence, we use mainly the direct current (dc) gm/ID sweeping curves in sizing to avoid fine tuning the parasitic effects on the frequency response.
- 3.
We adopt small-signal design equations (on gain, poles, zeros, GBW, phase margin, common-mode gain, etc.) together with other topologically derived design equations (on noise, common-mode range, and slew rate, etc.) to develop sizing constraints.
- 4.
We customize the ac design equations with the gm/ID variables to infer constraints on the gm/ID values from the design equations and the specification metrics.
- 5.
We also propose less expensive SPICE iteration methods to refine the improperly inferred sizing results by a few heuristic sizing rules.
Initial sizing by gm/ID for multi-stage Op Amps with the above listed features have not been completely addressed in a systematic way yet previously. A stage-by-stage initial sizing method is intuitive (i.e., totally based on circuit topologies), easy to follow by the beginners, and deterministic (requiring less SPICE iterations). More importantly, a sizing procedure so developed can be more easily adapted to automation once design equations can be automatically generated. The reader is referred to the works by Shi [9,10,33] for the recent progress.
This paper is organized as follows. In section 2 we discuss several commonly used amplifier stages and the gm/ID method currently well-known for sizing. Then in section 3 we develop procedural sizing methods for single-stage and multiple-stage amplifiers. Methods of how to customize the circuit-level design equations amenable to gm/ID constraints are discussed there, together with detailed sample designs on a one-stage, a two-stage, and a three-stage Op Amp. Finally, this paper is concluded in section 4.
Section snippets
Op Amp stages and gm/ID-based sizing
Multi-stage CMOS Op Amp designs have been extensively studied in the past decades, see the survey papers like [34,35]. A multi-stage Op Amp typically consists of several cascaded stages, each stage offering a partial dc gain. For achieving better frequency response properties like bandwidth (BW), GBW, and phase margin (PM), different kinds of frequency compensation strategies have been proposed in the literature [4,7,8,[36], [37], [38], [39], [40], [41], [42], [43], [44]]. Often, innovation of
Simulation setup
We present sizing of three Op Amps in this section to validate the gm/ID-based initial sizing method developed previously. The three Op Amps include a single-stage differential Op Amp shown in Fig. 2, a two-stage Op Amp shown in Fig. 10, and a three-stage Op Amp shown in Fig. 18, all having the same type of differential input stage. The two-stage Op Amp has a regular common source stage as the second stage and is compensated by a single Miller capacitor Cc (called the simple Miller
Conclusion
We have studied systematically the application of the gm/ID method in combination with design equations to initial sizing of multi-stage Op Amps. Since the main purpose of using a multi-stage Op Amp is to achieve a high dc gain while maintaining a good high frequency property to cope with a variety of driving needs, we have emphasized the gain, GBW, and PM as the foremost design metrics in our proposed sizing methodology. We have justified that several heuristic sizing tactics, such as the RDS
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (50)
- et al.
Systematic design and optimization of operational transconductance amplifier using gm/ID design methodology
Microelectron. J.
(2018) Toward automated reasoning for analog IC design by symbolic computation – a survey
Integration, the VLSI Journal
(2018)- et al.
Design three-stage dynamic-biased CMOS amplifier with a robust optimization of the settling time
IEEE Trans. Circuits and Systems–I: Regular Papers
(2015) - et al.
Optimized active single-Miller capacitor compensation with inner half-feedforward stage for very high-load three-stage OTAs
IEEE Trans. on Circuits and Systems–I: Regular Papers
(2016) - et al.
0.7-V three-stage class-AB CMOS operational transconductance amplifier
IEEE Trans. on Circuits and Systems–I: Regular Papers
(2016) - et al.
High-performance three-stage single-Miller CMOS OTA with no upper limit of CL
IEEE Trans. on Circuits and Systems–II: Express Briefs
(2018) - et al.
A 1.2-V 2.41-GHz three-stage CMOS OTA with efficient frequency compensation technique
IEEE Trans. on Circuits and Systems–I: Regular Papers
(2019) - et al.
In-depth analysis of pole-zero compensations in CMOS operational transconductance amplifiers
IEEE Trans. Circuits and Systems–I: Regular Papers
(2019) - et al.
A 0.016-mm2 144-μW three-stage amplifier capable of driving 1-to-15 nF capacitive load with 0.95-MHz GBW
IEEE J. Solid State Circ.
(2013) - et al.
Graphical analysis and design of multistage operational amplifiers with active feedback Miller compensation
Int. J. Circ. Theor. Appl.
(2016)
Topological approach to symbolic pole-zero extraction incorporating design knowledge
IEEE Trans. Comput. Aided Des. Integrated Circ. Syst.
Automatic stage-form circuit reduction for multistage opamp design equation generation
ACM Trans. Des. Autom. Electron. Syst.
Efficient hybrid performance modeling for analog circuits using hierarchical shrinkage priors
IEEE Trans. Comput. Aided Des. Integrated Circ. Syst.
An efficient Bayesian optimization approach for automated optimization of analog circuits
IEEE Transactions on Circuits and Systems – I: Regular Papers
Graph-constrained sparse performance modeling for analog circuit optimization via SDP relaxation
IEEE Trans. Comput. Aided Des. Integrated Circ. Syst.
Design of Analog CMOS Integrated Circuits
Design procedure for two-stage CMOS transconductance operational amplifiers: a tutorial
Analog Integr. Circuits Signal Process.
Design procedure for two-stage CMOS operational amplifiers employing current buffer
IEEE Trans. on Circuits and Systems–II: Express Briefs
Design methodology and advances in nested-Miller compensation
IEEE Trans. Circuits and Systems–I: Fundamental Theory and Applications
A gm/ID based methodology for the design of CMOS analog circuits and its application to the synthesis of a silicon-on-insulator micropower OTA
IEEE J. Solid State Circ.
Improved synthesis of gain-boosted regulated-cascode CMOS stages using symbolic analysis and gm/ID methodology
IEEE J. Solid State Circ.
Design optimization of high-speed and low-power operational transconductance amplifier using gm/ID lookup table methodology
IEICE Trans. Electron.
Systematic circuit design and analysis using generalised gm/ID functions of MOS devices
IET Circuits, Devices & Systems
Systematic Design of Analog CMOS Circuits Using Pre-computed Lookup Tables
DELIGHT.SPICE: An optimization-based system for the design of integrated circuits
IEEE Trans. on Computer-Aided Design
Cited by (9)
Application of improved ACM model to the design by hand of CMOS analog circuits
2022, Microelectronics JournalCitation Excerpt :Anyway, the by-hand design models usually do not take into account small dimension effects, such as charge sharing, carrier velocity saturation, drain induced barrier lowering (DIBL), neither the mobility degradation due to transversal field, otherwise equations should become cumbersome. The so-called gm/ID design methodology [12-18] is a different approach in which trade-off charts or lookup tables are generated from simulation to relate the transistor parameters with the transconductance-to-current ratio and the channel length. Since based on a wide simulated dataset, the second order effects in device operation are considered.
Multistage Op Amp design space exploration by g<inf>m</inf>/I<inf>D</inf> sampling and symbolic design equations
2024, International Journal of Circuit Theory and ApplicationsAutomatic generation of macromodels and design equations for application to Op Amp design
2023, International Journal of Circuit Theory and ApplicationsPrototype Design and Analysis of 4-Stage Operational-Amplifier using Double-Gate MOSFET
2023, International Journal of Engineering Trends and TechnologyDeep Neural Networks-Based Direct-Current Operation Prediction and Circuit Migration Design
2023, Electronics (Switzerland)Design and Implementation of Operational Amplifiers with CMOS 180 nm Technology Node using gm/ID Methodology
2023, 2023 IEEE 20th India Council International Conference, INDICON 2023
- ☆
This research was supported in part by the National Key R&D Program of China No. 2019YFB2205002 and the National Natural Science Foundation of China (NSFC) grant No. 61974087.