CN102289549A - Optimization method for analog integrated circuit - Google Patents
Optimization method for analog integrated circuit Download PDFInfo
- Publication number
- CN102289549A CN102289549A CN201110229280XA CN201110229280A CN102289549A CN 102289549 A CN102289549 A CN 102289549A CN 201110229280X A CN201110229280X A CN 201110229280XA CN 201110229280 A CN201110229280 A CN 201110229280A CN 102289549 A CN102289549 A CN 102289549A
- Authority
- CN
- China
- Prior art keywords
- posynomial
- design
- circuit
- optimization
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention belongs to the technical field of an integrated circuit design and particularly relates to an optimization method for an analog integrated circuit. The optimization method provided by the invention combines an optimization method based on formulas and an optimization method based on simulation; the performance parameters of the analog integrated circuit are expressed into positive term models about design variables, and an optimization design is carried out on the circuit by utilizing a geometric programming method; and furthermore, the positive term models in geometric programming problems are alternately corrected by utilizing an SPICE (simulation Program with Integrated Circuit) simulation result of a transistor level, thereby overcoming the defect of insufficient accuracy of the positive term models. In the method provided by the invention, the analog integrated circuit is optimized; the optimization accuracy of an SPICE level can be achieved in less computation time; and the problem that an existing method cannot give consideration to the optimization efficiency and the optimization accuracy is solved.
Description
Technical field
The invention belongs to integrated circuit (IC) design or Computer-aided Design Technology field, be specifically related to a kind of Analogous Integrated Electronic Circuits optimization method.
Background technology
Along with integrated circuit (Integrated Circuit IC) the enters the SoC(SOC (system on a chip)) epoch, the market demand of composite signal integrated circuits is increasing.The numerical portion of mixed-signal IC can be by mature C AD(computer-aided design (CAD)) instrument and and design methodology realize efficiently; Yet the simulation part is still main by manual design, makes it become the bottleneck of SoC design.Therefore, the IC design field presses for the CAD software that exploitation is applicable to the Analogous Integrated Electronic Circuits design, wherein the most important thing is synthesis tool.
Since late 1980s, the comprehensive main flow approach of Analogous Integrated Electronic Circuits is based on the method for optimization, promptly with design objective as constraint condition, adopt numerical optimization technique (mathematical programming etc.) to come the performance of optimized circuit.This integrated approach based on optimization generally includes two main modular: circuit performance assessment models and mathematical optimization engine.The former measured the circuit performance of current design point in each step of optimizing process; The latter adjusts design parameter according to the circuit performance knowledge that obtains, to optimize the performance of circuit.
According to the difference of the circuit performance assessment models that is adopted, can be divided into two big classes based on the integrated approach of optimizing: based on the method (1) (2) of formula, and based on method of emulation (3) (4).Adopt simple formula model to come the performance of evaluation circuits based on equation, shape as
(here
The expression phase margin,
,
,
Design variables such as the expression transistor is wide, long, bias current).Because the Performance Evaluation model is simple, and this explicit expression is easy to by the mathematical optimization programmed control, so these class methods have very high efficient.In recent years, some achievements in research find that the design object function of a large amount of Analogous Integrated Electronic Circuits and constraint condition can be expressed as (perhaps approximate representation is) posynomial (Posynomial) function about design parameter, therefore can adopt geometric programming (Geometric programming, GP) method comes circuit is optimized, and it has, and execution efficient is high, the characteristics (1) (2) of global optimization.Yet because simple formula model is difficult to accurately describe the circuit behavior of physics complexity, therefore the shortcoming based on the optimization method maximum of formula is the precision deficiency.
Corresponding with it, come the performance of evaluation circuits by the SPICE emulation (a kind of integrated circuit simulating instrument of industrial standard) of transistor level based on the optimization method of emulation.These class methods have the SPICE class precision as industrial standard, but owing to need carry out a large amount of SPICE emulation, they can bring great Computing burden.Typical case's representative of these class methods is DELIGHT.SPICE (3), and it is embedded into a SPICE emulator among the general optimum engine DELIGHT, carries out complete SPICE emulation in each step of optimizing iteration.DELIGHT.SPICE can only find the local optimum point, therefore is applicable to the circuit design that initial designs point is provided is carried out the part fine setting.Another one example based on emulation mode is ANACONDA (4), the behavior that it comes acquisition cuicuit with accurate SPICE emulation, and adopt evolution algorithm to seek optimum solution.ANACONDA can find globally optimal solution (being difficult to control in the reality) in theory, and supports parallel processing.Yet 50000-100000 SPICE emulation still can bring heavy computation burden altogether.
List of references
(1)?Hershenson?M,?Boyd?S,?Lee?T.?Optimal?design?of?a?cmos?op-amp?via?geometric?programming.?IEEE?Transactions?on?Computer-Aided?Design?of?Integrated?Circuits?and?Systems,?2001,?20(1):?1-21
(2)?Mandal?P,?Visvanathan?V.?CMOS?op-amp?sizing?using?a?geometric?programming?formulation.?IEEE?Transactions?on?Computer-Aided?Design?of?Integrated?Circuits?and?Systems,?2001,?20(1):?22-38
(3)?Nye?W,?Riley?D,?Sangiovanni-Vincentelli?A,?Tits?A.?Delight.spice:?an?optimization-based?system?for?the?design?of?integrated?circuits.?IEEE?Transactions?on?Computer-Aided?Design?of?Integrated?Circuits?and?Systems,?1988,?7(4):?501-519
(4)?Phelps?R,?Krasnicki?M,?Rutenbar?R?A,?et?al.?Anaconda:?Simulation-based?synthesis?of?analog?circuits?via?stochastic?patter?search.?IEEE?Transactions?on?Computer-Aided?Design?of?Integrated?Circuits?and?Systems,?2000,?19(6):?703-717.。
Summary of the invention
The objective of the invention is to overcome the problem that existing optimization method can't be taken into account precision and efficient simultaneously, proposing a kind of existing very high precision has very high efficiency Analogous Integrated Electronic Circuits optimization method again.
The Analogous Integrated Electronic Circuits optimization method that the present invention proposes combines based on equation and advantage based on simulation method: with the SPICE simulation result formula model of correcting circuit performance iteratively, to improve the precision based on equation.Use this method Analogous Integrated Electronic Circuits is optimized design, can in computing time seldom, reach the optimization precision of SPICE level.
The present invention is a kind of in conjunction with geometric programming (Geometric Programming GP) with the Analogous Integrated Electronic Circuits optimization method of model tuning (Model Modification), abbreviates MMGP(Model Modified Geometric Programming as).The framed structure of MMGP as shown in Figure 1, it realizes concrete steps following (as shown in Figure 2):
Step 1 is expressed as posynomial form about design parameter, i.e. initialization posynomial model with the objective function of circuit design and constraint condition;
Step 3 utilizes posynomial model and SPICE emulator to remove to assess design parameter
The performance of following circuit promptly obtains respectively
And
The former represents posynomial model evaluation result, and the latter represents the SPICE simulation result;
By repeating step 2-4, MMGP carries out the process of " model tuning-optimal design " iteratively, finally converges on an optimal design point with SPICE class precision.
Below from three aspects realization details of the present invention is described respectively: the foundation of posynomial model, based on the correction of the circuit optimization and the posynomial model of geometric programming.
1, the foundation of posynomial model:
In the present invention, initial posynomial model is derived by the experimental formula of Analogous Integrated Electronic Circuits design and is obtained.Experimental formula is commonly used to calculate initial designs point and provides reference intuitively for deviser's debug circuit.Yet these experimental formulas are expressed as the performance of circuit the function of small-signal variable usually, and for example (phase margin PM) is typically expressed as the phase margin of operational amplifier
(wherein
,
,
Represent transistorized mutual conductance, output resistance and grid source electric capacity) form.Therefore, at first need to convert these experimental formulas to direct function, for example about design parameter
Form.Then, with original design object (for example:
) and design constraint is (for example
, here
Represent corresponding design objective) be converted to GP(geometric programming) compatible form (
,
), require here corresponding expression (
,
) be posynomial.Subordinate list 1 has provided various design constraint or how design object converts the form of GP compatibility to, and has listed corresponding posynomial requirement.
Fortunately, the design object function of a large amount of Analogous Integrated Electronic Circuits and constraint function can natural surface be shown as the posynomial form into design parameter, perhaps can take some approximation methods to be converted to the posynomial function.
2, based on the circuit optimization of geometric programming:
After design object function and design constraint function table be shown as the posynomial form, can become the initial optimization problem description of circuit the form of geometric programming easily:
Wherein
Be optimization variable, the design parameter of indication circuit here, promptly
, here
,
,
,
,
Represent transistorized wide, long, bias current, resistance, and electric capacity respectively;
Expression design object function has the form of posynomial;
(
) expression Performance Constraints condition (design objective), have the form of posynomial;
(
) and
(
) representative biasing constraint condition (between transistor operationg region, pipe coupling require etc.):
Be posynomial,
Be single posynomial (having and have only one posynomial);
The expression technology is to the restriction of the size of circuit devcie, and wherein, the former is a lower limit, and the latter is the upper limit;
,
,
The number of representing corresponding constraint condition respectively.Geometric programming problem (1) can be converted to a protruding optimization problem, therefore can try to achieve globally optimal solution with high efficient.
3, posynomial model tuning:
MMGP proofreaies and correct the posynomial model by a kind of mode very intuitively: multiply by correction factor and (annotate: in the iterative process on initial posynomial model based, it all is to multiply by correction factor on initial posynomial model based that each step proofreaies and correct), and the ratio of the performance parameter that circuit performance parameters that correction factor is obtained by SPICE emulation and posynomial model evaluation obtain obtains.In this way, each step iterative process only needs to carry out a SPICE emulation, so just the time cost of model tuning can be dropped to minimum.And the posynomial model multiply by the form that positive correction factor can't change posynomial, still can optimize apace with original geometric programming method.After the model tuning, geometric programming problem (1) originally is corrected for following form:
Wherein
Expression is at
The correction factor of item performance parameter is calculated as follows:
(3)
Wherein
The optimal design point that expression previous step iteration obtains;
With
The expression design parameter is
The time circuit
The item performance parameter: the former is the SPICE simulation result, and the latter is posynomial model evaluation result.
Last item constraint condition of geometric programming problem (2)
Meaning be hunting zone with geometric programming be restricted to previous step iteration gained optimization result (
) a neighborhood in (
), here
Be called compressibility factor.And along with the carrying out of iteration, the hunting zone that MMGP can reduce geometric programming gradually (promptly reduces compressibility factor gradually
).The precision of model tuning can be progressively improved like this, also convergence of algorithm can be guaranteed.The value of compressibility factor and the mode that reduces are determined on a case-by-case basis, in the specific embodiment of this paper, and compressibility factor
Initial value be
, whenever carry out single-step iteration,
Value reduce by half.
Compare with existing Analogous Integrated Electronic Circuits optimization method, MMGP has taken into account precision and efficient well.Compare based on the circuit optimization method of formula with tradition, MMGP has the precision of SPICE level, and the cost of for this reason paying is the overhead that a small amount of SPICE emulation of operation brings.Compare with the optimization method method based on emulation that has the SPICE class precision equally, MMGP has absolute advantage on efficient.Subordinate list 2 is for two-stage miller compensation amplifier, the comparison sheet of the circuit performance that MMGP method of the present invention and the optimization of conventional geometric planing method obtain.
Description of drawings
Fig. 1 is the structural representation of Analogous Integrated Electronic Circuits optimization method of the present invention.
Fig. 2 is the algorithm flow chart of Analogous Integrated Electronic Circuits optimization method of the present invention.
Fig. 3 is the circuit structure diagram of the specific embodiment (two-stage miller compensation amplifier) of Analogous Integrated Electronic Circuits optimization method of the present invention.
Fig. 4 is for two-stage miller compensation amplifier, and MMGP method of the present invention is at the convergence process figure in circuit performance space.
Embodiment
For above-mentioned purpose of the present invention, feature and advantage can be become apparent more, further specify the present invention below by a concrete example.
One embodiment of the present of invention are that the two-stage miller compensation operation amplifier circuit shown in the accompanying drawing 3 is optimized.This circuit is made up of biasing circuit, differential input stage and common source amplifier stage; Resistance
And electric capacity
Be used for phase compensation.Two-stage miller compensation amplifier shown in the accompanying drawing 3 has 18 design parameters: 8 transistorized length and widths (
), bias current
, and building-out capacitor
Compensating resistance
Value by the decision of other design parameters, promptly
, here
The expression transistor
Mutual conductance.
The design adopts
The CMOS(complementary metal oxide semiconductor (CMOS)) technology; We are with supply voltage
,
Be fixed as respectively
,
Load capacitance is fixed as
Design object is in the unity gain bandwidth that satisfies maximization amplifier under the condition of design constraint.Initial geometric programming problem description is as follows:
Optimization aim(unity gain bandwidth GBW):
Optimize constraint:
D. (index is open-loop gain
)
J. biasing constraint
K. the process technology limit of design parameter
In above optimization problem,
, wherein
,
The gate oxide electric capacity of representing hole mobility, unit area respectively;
, wherein
The expression hole mobility;
,
The threshold voltage of representing PMOS pipe (P-type mos transistor) and NMOS pipe (N type metal oxide semiconductor transistor) respectively;
,
The channel length modulation coefficient of representing PMOS pipe and NMOS pipe respectively;
,
The noise figure of representing PMOS pipe and NMOS pipe respectively;
The expression Boltzmann constant;
The expression kelvin degree;
The indication circuit frequency of operation is got here
Here adopt the emulated data of HSPICE emulation tool Level 1 model the posynomial model in the above-mentioned geometric programming problem to be proofreaied and correct in each step of algorithm iteration.(annotate: HSPICE is one of current most popular SPICE emulation tool for the commercial SPICE emulation tool of Synopsys company exploitation.)
The data that are labeled as " MMGP " in the subordinate list 2 have been listed the performance parameter of the optimal design that is obtained by the MMGP method (HSPICE emulated data).In the present example, the MMGP algorithm has carried out iteration altogether 10 times, that is to say that whole optimizing process only need move 10 HSPICE emulation.From the optimization result of MMGP method as can be seen, all properties parameter of this design all satisfies design objective; In addition, the performance parameter and the design objective of responsive bound term (phase margin, quiescent dissipation, equivalent input noise) accurately mate, and have avoided crossing design.Therefore, MMGP has improved the precision (reached the optimization precision of HSPICE level) of tradition based on the circuit optimization method of GP greatly, and the cost of for this reason paying is to carry out expense extra time that small amount of H SPICE emulation brings.
Accompanying drawing 4 has write down the iterative process of MMGP method optimization two stage amplifer from performance space by some examples (quiescent dissipation, phase margin, equivalent input noise, unity gain bandwidth).As can be seen from the figure MMGP method speed of convergence is exceedingly fast, and has verified its efficient.
Subordinate list 1
Former design constraint (target) | GP design constraint (target) | The posynomial requirement |
Be posynomial | ||
Be posynomial | ||
Be posynomial | ||
Be posynomial | ||
Be single posynomial |
(annotate: go up in the table
(comprise
,
) objective function or the constraint function of indication circuit design,
Represent corresponding design objective).
Claims (3)
1. an Analogous Integrated Electronic Circuits optimization method is characterized in that abbreviating MMGP as in conjunction with geometric programming and model tuning, and concrete steps are as follows:
Step 1 is expressed as posynomial form about design parameter, i.e. initialization posynomial model with the objective function of circuit design and constraint condition;
Step 2 uses the method for geometric programming that circuit is optimized, and obtains the substandard optimum circuit design parameter of posynomial model evaluation
Step 3 utilizes posynomial model and SPICE emulator to remove to assess design parameter
The performance of following circuit obtains respectively
And
, the former represents posynomial model evaluation result, the latter represents the SPICE simulation result;
Step 4 is according to the Different Results of two kinds of assessment modes
And
, remove to proofread and correct initial posynomial model, to improve model accuracy;
By repeating step 2-4, carry out the process of " model tuning-optimal design " iteratively, finally converge on an optimal design point with SPICE class precision.
2. Analogous Integrated Electronic Circuits optimization method according to claim 1, after it is characterized in that circuit design objective function and design constraint function table be shown as the posynomial form, the initial optimization problem description of circuit becomes the form of geometric programming as follows:
Wherein
Be optimization variable, the design parameter of indication circuit here, promptly
, here
,
,
,
,
Represent transistorized wide, long, bias current, resistance, and electric capacity respectively;
Expression design object function has the form of posynomial;
Expression Performance Constraints condition has the form of posynomial,
And
Representative biasing constraint condition,
Be posynomial,
Be single posynomial,
J=1
..., n,
The expression technology is to the restriction of the size of circuit devcie, and wherein, the former is a lower limit, and the latter is the upper limit;
,
,
The number of representing corresponding constraint condition respectively.
3. Analogous Integrated Electronic Circuits optimization method according to claim 2 is characterized in that after the model tuning, and geometric programming problem (1) originally is corrected for following form:
Wherein,
Expression is at
The correction factor of item performance parameter is calculated as follows:
In the formula,
The optimal design point that expression previous step iteration obtains;
With
The expression design parameter is
The time circuit
The item performance parameter: the former is the SPICE simulation result, and the latter is posynomial model evaluation result, in the formula (2)
The expression compressibility factor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110229280XA CN102289549A (en) | 2011-08-11 | 2011-08-11 | Optimization method for analog integrated circuit |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201110229280XA CN102289549A (en) | 2011-08-11 | 2011-08-11 | Optimization method for analog integrated circuit |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102289549A true CN102289549A (en) | 2011-12-21 |
Family
ID=45335974
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201110229280XA Pending CN102289549A (en) | 2011-08-11 | 2011-08-11 | Optimization method for analog integrated circuit |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102289549A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105303008A (en) * | 2015-12-03 | 2016-02-03 | 中国科学院微电子研究所 | Method and system for optimizing analogue integrated circuit |
CN110245436A (en) * | 2019-06-19 | 2019-09-17 | 山东大学 | A kind of Parallel Simulation circuit optimization method based on genetic algorithm and machine learning |
CN112417803A (en) * | 2020-12-02 | 2021-02-26 | 苏州复鹄电子科技有限公司 | Artificial intelligence algorithm-based automatic optimization scheme for design parameters of analog integrated circuit |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101464919A (en) * | 2008-12-30 | 2009-06-24 | 上海集成电路研发中心有限公司 | BSIM3 HCI reliability model used in MOSFET electrical simulation |
US20100299649A1 (en) * | 2004-09-30 | 2010-11-25 | Mar Hershenson | Novel optimization for circuit design |
-
2011
- 2011-08-11 CN CN201110229280XA patent/CN102289549A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100299649A1 (en) * | 2004-09-30 | 2010-11-25 | Mar Hershenson | Novel optimization for circuit design |
CN101464919A (en) * | 2008-12-30 | 2009-06-24 | 上海集成电路研发中心有限公司 | BSIM3 HCI reliability model used in MOSFET electrical simulation |
Non-Patent Citations (2)
Title |
---|
ZENG XUAN等: "《Optimal Stack Generation for CMOS Analog Modules with Parasitic and Mismatch Constraints》", 《半导体学报》 * |
蒋华,戎蒙恬,刘文江: "《基于几何规划的模拟器件建模与电路性能优化》", 《信息技术》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105303008A (en) * | 2015-12-03 | 2016-02-03 | 中国科学院微电子研究所 | Method and system for optimizing analogue integrated circuit |
CN105303008B (en) * | 2015-12-03 | 2019-02-05 | 中科芯云微电子科技有限公司 | A kind of optimization method for analog integrated circuit and system |
CN110245436A (en) * | 2019-06-19 | 2019-09-17 | 山东大学 | A kind of Parallel Simulation circuit optimization method based on genetic algorithm and machine learning |
WO2020253055A1 (en) * | 2019-06-19 | 2020-12-24 | 山东大学 | Parallel analog circuit optimization method based on genetic algorithm and machine learning |
CN112417803A (en) * | 2020-12-02 | 2021-02-26 | 苏州复鹄电子科技有限公司 | Artificial intelligence algorithm-based automatic optimization scheme for design parameters of analog integrated circuit |
CN112417803B (en) * | 2020-12-02 | 2024-02-06 | 苏州复鹄电子科技有限公司 | Automatic optimization method for design parameters of analog integrated circuit based on artificial intelligence algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7761820B2 (en) | Automated migration of analog and mixed-signal VLSI design | |
Maulik et al. | Sizing of cell-level analog circuits using constrained optimization techniques | |
Seth et al. | Settling time and noise optimization of a three-stage operational transconductance amplifier | |
CN103150459B (en) | A kind of Optimization Design of the low noise amplifier based on genetic algorithm | |
Li et al. | Robust analog/RF circuit design with projection-based performance modeling | |
CN104007777A (en) | Current source generator | |
Graupner et al. | Generator based approach for analog circuit and layout design and optimization | |
CN102289549A (en) | Optimization method for analog integrated circuit | |
Iskander et al. | Hierarchical sizing and biasing of analog firm intellectual properties | |
Liao et al. | An LDE-Aware g m/I D-Based Hybrid Sizing Method for Analog Integrated Circuits | |
Liao et al. | Layout-dependent effects aware g m/i D-based many-objective sizing optimization for analog integrated circuits | |
Maji et al. | An evolutionary approach based design automation of low power CMOS Two-Stage Comparator and Folded Cascode OTA | |
Dolatshahi et al. | A new systematic design approach for low-power analog integrated circuits | |
Fortes et al. | Low power bulk-driven OTA design optimization using cuckoo search algorithm | |
Lourenço et al. | Analog circuit design based on robust POFs using an enhanced MOEA with SVM models | |
CN109117528B (en) | MOS device subcircuit temperature model based on BSIM4 model and modeling method | |
Guo | An accurate design approach for two-stage CMOS operational amplifiers | |
Girardi et al. | Power constrained design optimization of analog circuits based on physical gm/ID characteristics | |
JP2008181377A (en) | System and method for designing semiconductor device | |
Kourany et al. | Passiot: a pareto-optimal multi-objective optimization approach for synthesis of analog circuits using sobol’indices-based engine | |
Girardi et al. | A tool for automatic design of analog circuits based on gm/i/sub d/methodology | |
Lv et al. | Exact design for the settling time of two-stage Miller compensated operational amplifiers | |
CN105760593A (en) | MOS transistor NBTI effect R-D model parameter extraction method | |
Dam et al. | Iterative performance model upgradation in geometric programming based analog circuit sizing for improved design accuracy | |
Cashero et al. | Fast evaluation of analog circuits using linear programming |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20111221 |