## Statistically based parametric yield prediction for integrated circuits (1997)

Venue: | IEEE Transactions On Semiconductor Manufacturing |

Citations: | 8 - 2 self |

### BibTeX

@ARTICLE{Gibson97statisticallybased,

author = {David S. Gibson and Ravi Poddar and Gary S. May and Senior Member and Martin A. Brooke},

title = {Statistically based parametric yield prediction for integrated circuits},

journal = {IEEE Transactions On Semiconductor Manufacturing},

year = {1997},

volume = {10},

pages = {445--458}

}

### Years of Citing Articles

### OpenURL

### Abstract

Abstract—This paper presents a novel procedure for predicting integrated circuit parametric performance and yield when provided with sample transistor test results and a circuit schematic. Two enhancements to the existing Monte Carlo simulation procedures are described: 1) a multivariate nested model is used to reproduce random process-induced device variations, rather than the multivariate multinormal model typically used, and 2) the stochastic Monte Carlo method for mapping process variability into a performance distribution is replaced with a deterministic mapping technique. The use of multivariate nested distributions allows estimation not only of correlation between various model parameters, but also allows each of those variations to be apportioned among the various stages of the process (i.e., wafer to wafer, lot to lot, etc.). This allows matched devices to be more accurately simulated, without having to develop customized models for each configuration of matching, and provides focus for process improvement efforts into those areas with the maximum potential reward. The use of deterministic mapping provides simulation results which are repeatable and do not rely on chance to insure that the process parameter space has been evenly explored. A software package which implements the entire procedure has been written in C++. Index Terms—Monte Carlo simulation, multivariate statistics, parametric yield.

### Citations

639 |
Design and analysis of experiments
- Montgomery
(Show Context)
Citation Context ...for guiding the selection of Monte Carlo instances to improve the accuracy of yield estimates for a fixed sample size. Input variables are often examined using analysis of variance (ANOVA) techniques =-=[2]-=- to eliminate statistically insignificant parameters [3], [4]. Uncorrelated multinormal distributions are often used, with many authors choosing the same four supposedly independent parameters for MOS... |

539 |
A comparison of three methods for selecting values of input variables in the analysis of output from a computer code
- Mckay, Beckman, et al.
- 1979
(Show Context)
Citation Context ...29] performs a similar function, but includes elements of stratified sampling as well. Latin hypercube sampling insures that a wide range of values are used for each variable in a multivariate sample =-=[30]-=-. Given this history, there is room for improvement in at least three areas: the simulation of analog circuits which use matched devices, extracting information from the process model to maximize the ... |

461 |
Monte Carlo Methods
- Hammersley, Handscornb
- 1964
(Show Context)
Citation Context ...rated circuits, modify fabrication processes, plan production schedules, or specify product operating characteristics. Typically, this prediction is accomplished using a three-step Monte Carlo method =-=[1]-=-: 1) a statistical model is built to characterize the fabrication process to be used; 2) a circuit design is created using a circuit simulator (such as SPICE) and nominal device values for the target ... |

428 |
Applied linear statistical models
- Neter, MH, et al.
- 1996
(Show Context)
Citation Context ...n. Nested distributions are appropriate when the phenomena being modeled can be divided into “stages” such that the variability within one stage is independent of the variability within another stage =-=[34]-=-. In semiconductor manufacturing, the potential stages are processes, lots, wafers, circuits, groups (of matched devices), and devices. These are generally mutually independent in their variability. F... |

29 |
A Symbolic Simulator for Analog Integrated Circuits
- Gielen, Sansen
- 1989
(Show Context)
Citation Context ...ted using linear [7], piecewise linear [16], quadratic [17], [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used =-=[20]-=-, as can hierarchical model generation [4], [21], [22], [23]. Simplicial approximation attempts to reduce the need for simulation by identifying the acceptance region in input parameter space [24]. Se... |

23 | Predictive Subset Testing: Optimizing IC Parametric Performance for Quality, Cost and Yield - Brockman, Director - 1989 |

10 | Parametric yield optimization for MOS circuit blocks - Hocevar, Cox, et al. - 1988 |

10 |
A Study of Variance Reduction Techniques for Estimating Circuit Yields
- Hocevar, Lightner, et al.
- 1983
(Show Context)
Citation Context ...ng the accuracy of a Monte Carlo yield estimate without increasing sample size. Each is dependent on a controlled alteration of the random generation of the input instances. Importance sampling [25], =-=[26]-=- forces more samples in regions where yield is close to 50%. Parametric sampling [27] allows samples froms446 IEEE TRANSACTIONS ON SEMICONDUCTOR MANUFACTURING, VOL. 10, NO. 4, NOVEMBER 1997 previous c... |

9 |
Statistical modeling for efficient parametric yield estimation
- Cox, Yang, et al.
- 1983
(Show Context)
Citation Context ...ated multinormal distributions are often used, with many authors choosing the same four supposedly independent parameters for MOSFET devices: length, width, gate oxide thickness, and flatband voltage =-=[5]-=-–[10]. Others have preferred to use manufacturing parameters such as diffusion times and temperatures [3], [11]. Often, correlation is accounted for through the use of principal component analysis [12... |

6 |
Signal integrity analysis and optimization of VLSI interconnects using neural network models
- Zhang, Nakhla
- 1994
(Show Context)
Citation Context ...d, it is desirable to minimize the calculation time for each simulation. The data resulting from a handful of well chosen SPICE simulations can be used to develop a polynomial or neural network model =-=[14]-=-, [15], which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear [16], quadratic [17], [1... |

6 |
DOE/Opt: A System for Design of Experiments, Response Surface Modeling, and Optiminzation Using Process and Device Simulation
- Boning, Mozumder
- 1994
(Show Context)
Citation Context ...can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear [16], quadratic [17], [18], and cubic =-=[19]-=- models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], [21], [22], [23]. Simplicial appr... |

6 |
Quality Assurance and Upgrade of Analog Characteristics by Fast Mismatch Analysis Option in Network Analysis Environment
- Oehm, Schumacher
- 1993
(Show Context)
Citation Context ...hey exhibit. The second method develops an empirical model for each specific matched device configuration, and uses that model to generate specific parameter information for each matched device [32], =-=[33]-=-. This technique requires explicit model development for each new matching configuration (for example, if three matching devices were used rather than two), and assumes that the only within-die variat... |

5 |
Statistical modeling of VLSI circuit performances
- Yu, Kang, et al.
- 1986
(Show Context)
Citation Context ...elop a polynomial or neural network model [14], [15], which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear =-=[7]-=-, piecewise linear [16], quadratic [17], [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarc... |

4 |
A Method for Modeling the Manufacturability of IC Designs
- Boskin, Spanos, et al.
- 1994
(Show Context)
Citation Context ...0]. Others have preferred to use manufacturing parameters such as diffusion times and temperatures [3], [11]. Often, correlation is accounted for through the use of principal component analysis [12], =-=[13]-=-. Given a generated circuit instance, many techniques have been used to predict the resulting performance parameters. The most common of these, and the standard by which the accuracy of most technique... |

4 |
Acceptance sampling: an efficient, accurate method for estimating and optimizing parametric yield
- Elias
- 1994
(Show Context)
Citation Context ...yield in each region. The control variate technique [26], [28] can be used to overcome inaccuracies in the mapping function (typically employed only when non-SPICE maps are used). Acceptance sampling =-=[29]-=- performs a similar function, but includes elements of stratified sampling as well. Latin hypercube sampling insures that a wide range of values are used for each variable in a multivariate sample [30... |

4 |
PYFS—A statistical optimization method for integrated circuit yield enhancement
- Pan, Hu
- 1993
(Show Context)
Citation Context ...alues, then unrealistically good performance is predicted. Two approaches have attempted to rectify this problem. The first simulates within-die variations by adding a random component to each device =-=[31]-=-. This falls short in that it attempts to use the same variability for all devices in a design, regardless of the degree of matching they exhibit. The second method develops an empirical model for eac... |

4 |
Rapid yield estimation as a computer aid for analog circuit design
- Mukherjee, Carley
- 1991
(Show Context)
Citation Context ...hing they exhibit. The second method develops an empirical model for each specific matched device configuration, and uses that model to generate specific parameter information for each matched device =-=[32]-=-, [33]. This technique requires explicit model development for each new matching configuration (for example, if three matching devices were used rather than two), and assumes that the only within-die ... |

3 |
Analog statistical simulation
- Rencher
- 1991
(Show Context)
Citation Context ...prove the accuracy of yield estimates for a fixed sample size. Input variables are often examined using analysis of variance (ANOVA) techniques [2] to eliminate statistically insignificant parameters =-=[3]-=-, [4]. Uncorrelated multinormal distributions are often used, with many authors choosing the same four supposedly independent parameters for MOSFET devices: length, width, gate oxide thickness, and fl... |

3 |
Hierarchical yield estimation of large analog integrated circuits
- Kurker, Paulos, et al.
- 1993
(Show Context)
Citation Context ... the accuracy of yield estimates for a fixed sample size. Input variables are often examined using analysis of variance (ANOVA) techniques [2] to eliminate statistically insignificant parameters [3], =-=[4]-=-. Uncorrelated multinormal distributions are often used, with many authors choosing the same four supposedly independent parameters for MOSFET devices: length, width, gate oxide thickness, and flatban... |

3 |
Projection of Circuit Performance Distributions by Multivariate Statistics
- Chow
- 1989
(Show Context)
Citation Context ...[5]–[10]. Others have preferred to use manufacturing parameters such as diffusion times and temperatures [3], [11]. Often, correlation is accounted for through the use of principal component analysis =-=[12]-=-, [13]. Given a generated circuit instance, many techniques have been used to predict the resulting performance parameters. The most common of these, and the standard by which the accuracy of most tec... |

3 |
Statistical and Behavioral Modeling of Analogue Integrated
- Koskinen, Cheung
- 1993
(Show Context)
Citation Context ...dratic [17], [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], =-=[21]-=-, [22], [23]. Simplicial approximation attempts to reduce the need for simulation by identifying the acceptance region in input parameter space [24]. Several techniques are available for improving the... |

3 |
Computationally efficient yield estimation procedures based on simplicial approximation
- Director, Hatchel, et al.
- 1978
(Show Context)
Citation Context ...sed [20], as can hierarchical model generation [4], [21], [22], [23]. Simplicial approximation attempts to reduce the need for simulation by identifying the acceptance region in input parameter space =-=[24]-=-. Several techniques are available for improving the accuracy of a Monte Carlo yield estimate without increasing sample size. Each is dependent on a controlled alteration of the random generation of t... |

2 | Statistical device characterization and parametric yield estimation,” Solid State Technol - Cox, Yang, et al. - 1985 |

2 |
A macromodeling based approach for efficient IC yield optimization
- Feldman, Director
- 1991
(Show Context)
Citation Context ... multinormal distributions are often used, with many authors choosing the same four supposedly independent parameters for MOSFET devices: length, width, gate oxide thickness, and flatband voltage [5]–=-=[10]-=-. Others have preferred to use manufacturing parameters such as diffusion times and temperatures [3], [11]. Often, correlation is accounted for through the use of principal component analysis [12], [1... |

2 |
Predictive worst case statistical modeling of 0.8-"m BICMOS bipolar transistors: A methodology based on process and mixed device/circuit level simulators
- Kizilyalli, Ham, et al.
- 1993
(Show Context)
Citation Context ... parameters for MOSFET devices: length, width, gate oxide thickness, and flatband voltage [5]–[10]. Others have preferred to use manufacturing parameters such as diffusion times and temperatures [3], =-=[11]-=-. Often, correlation is accounted for through the use of principal component analysis [12], [13]. Given a generated circuit instance, many techniques have been used to predict the resulting performanc... |

2 |
An efficient tolerance design procedure for yield maximization using optimization techniques and neural network
- Chen, Chan
- 1991
(Show Context)
Citation Context ...is desirable to minimize the calculation time for each simulation. The data resulting from a handful of well chosen SPICE simulations can be used to develop a polynomial or neural network model [14], =-=[15]-=-, which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear [16], quadratic [17], [18], an... |

2 |
Parametric yield estimation for a MOSFET integrated circuit
- Berrah
- 1990
(Show Context)
Citation Context ...eural network model [14], [15], which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear =-=[16]-=-, quadratic [17], [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation ... |

2 |
Parametric yield optimization of CMOS analogue circuits by quadratic statistical circuit performance models,” Int
- Yu, Kang, et al.
- 1991
(Show Context)
Citation Context ...del [14], [15], which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear [16], quadratic =-=[17]-=-, [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], [21], [22],... |

2 |
An efficient method for parametric yield optimization of MOS integrated circuits
- Yu, Kang, et al.
- 1989
(Show Context)
Citation Context ...4], [15], which can then be used in place of SPICE for future, faster calculations. Polynomial circuit models have been successfully generated using linear [7], piecewise linear [16], quadratic [17], =-=[18]-=-, and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], [21], [22], [23].... |

2 |
Statistical design centering and tolerance using parametric sampling
- Singhal, Pinel
- 1981
(Show Context)
Citation Context ...is dependent on a controlled alteration of the random generation of the input instances. Importance sampling [25], [26] forces more samples in regions where yield is close to 50%. Parametric sampling =-=[27]-=- allows samples froms446 IEEE TRANSACTIONS ON SEMICONDUCTOR MANUFACTURING, VOL. 10, NO. 4, NOVEMBER 1997 previous circuit designs to be added to samples for more recent versions. Stratified sampling [... |

2 |
Efficient tolerance analysis using control variates,” Proc
- Soin, Rankin
- 1985
(Show Context)
Citation Context ...or more recent versions. Stratified sampling [26] divides the process region into a number of nonoverlapping regions, and then calculates the yield in each region. The control variate technique [26], =-=[28]-=- can be used to overcome inaccuracies in the mapping function (typically employed only when non-SPICE maps are used). Acceptance sampling [29] performs a similar function, but includes elements of str... |

2 |
received the
- London, Hall
- 1992
(Show Context)
Citation Context ...inations of distribution model and instance generation technique. (a) (b) (c) (d) Fig. 9. Output performance for the circuit instances shown in Fig. 8. through the use of principal component analysis =-=[36]-=-), then this procedure can simply be repeated once for each dimension. Creating a two-dimensional pdf using the distribution in Fig. 7 would result in the same five points for both axes, with the resu... |

1 |
Hierarchical tolerance using behavioral models
- Koskinen, Cheung
- 1992
(Show Context)
Citation Context ... [17], [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], [21], =-=[22]-=-, [23]. Simplicial approximation attempts to reduce the need for simulation by identifying the acceptance region in input parameter space [24]. Several techniques are available for improving the accur... |

1 |
Modeling behavior and tolerances in analogue cells
- Koskinen, Cheung
- 1991
(Show Context)
Citation Context ... [18], and cubic [19] models. Transformations are often made to improve the accuracy of these models. Analytic expressions can also be used [20], as can hierarchical model generation [4], [21], [22], =-=[23]-=-. Simplicial approximation attempts to reduce the need for simulation by identifying the acceptance region in input parameter space [24]. Several techniques are available for improving the accuracy of... |