CN104679960B - A kind of statistical modeling method of radio frequency variodenser - Google Patents
A kind of statistical modeling method of radio frequency variodenser Download PDFInfo
- Publication number
- CN104679960B CN104679960B CN201510111705.5A CN201510111705A CN104679960B CN 104679960 B CN104679960 B CN 104679960B CN 201510111705 A CN201510111705 A CN 201510111705A CN 104679960 B CN104679960 B CN 104679960B
- Authority
- CN
- China
- Prior art keywords
- parameter
- statistical
- radio frequency
- value
- variodenser
- 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.)
- Active
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing Or Measuring Of Semiconductors Or The Like (AREA)
Abstract
The statistical modeling method of the radio frequency variodenser of the present invention, including：Determine that organs weight needs all model parameters extracted；Parameter is classified；Parameter testing is scattered to device, establishes the benchmark model not comprising technological fluctuation of device；Wave characteristic of the device scattering parameter in whole wafer is obtained, it is determined that fluctuation data division；Based on the network parameter for fluctuating the scattering parameter of data division and being converted by scattering parameter, the index that device carries out statistical modeling is configured to；Based on the index, each component value in sub-circuit model is calculated, sensitivity analysis and sequence of the units value of going forward side by side relative to desired value, chooses sensitivity highest element；To all kinds of populations of parameters in the scalable formula of element, sensitivity analysis and sequence of the parameter value for component value are carried out respectively；Sensitivity highest parameter is chosen, is fitted the statistical property distribution of device, obtains the statistical model of the global fluctuation distribution of device.The modeling process of the present invention is quickly clear.
Description
Technical field
The present invention relates to semiconductor devices modeling technique technical field, and in particular to a kind of statistical modeling of radio frequency variodenser
Method.
Background technology
Variodenser (Varactor) has critically important as main capacitance adjustment device in RF IC design
Effect, with the reduction of process node, its performance inconsistency as caused by technique on full wafer wafer is also more and more notable.It is and current
The not disclosed report of modeling for the statistical property of variodenser radio-frequency performance.
The radio frequency modeling method of variodenser does not form unified standard, and different sub-circuit knots can be taken by showing as model
Structure, and the element in circuit can also use different scalability (Scalable) formula.And most formula do not have strictly
Physical significance, but the fitting based on mathematics, cause model parameter can not accurate respective devices fabrication process parameters.This is
Choose and bring difficulty for characterizing the statistical model parameter of technological fluctuation.
For different foundries (foundry) technique, its technological fluctuation reason is different, i.e., is shown as in test data
Different statistical distribution characteristic, it is difficult to choose the statistical modeling that unified fixed standard index carries out variodenser.
Therefore it provides a kind of statistical modeling method of the radio-frequency capacitor of simple possible seems particularly significant.
The content of the invention
In order to overcome problem above, the present invention is intended to provide a kind of statistical modeling method of radio frequency variodenser, to establish
For the global statistics model of the variodenser of RF IC, the influence of the device performance caused by technological fluctuation is characterized,
So as to make guidance to RF IC design.
To achieve these goals, the invention provides a kind of statistical modeling method of radio frequency variodenser, the radio frequency to become
Container is distributed on the crystal grain of wafer, and it comprises the following steps：
Step 01：It is determined that the sub-circuit topological structure for organs weight, determines each circuit elements in the sub-circuit
The scalability formula that part uses, for the modeling of different scale devices, so that it is determined that organs weight need to extract it is all
Model parameter；
Step 02：Parameter is classified, is divided into basic parameter group, the population of parameters related to applied bias, with device chi
Very little relevant parameter group, and the population of parameters unrelated with technological fluctuation；
Step 03：Based on radio frequency modeling demand, parameter testing, first selected section critical size device are scattered to device
Part is scattered the mapping tests of parameter, i.e., is all scattered parameter testing to the selected scale device in all crystal grains, point
The mapping test datas of scattering parameter are analysed, obtain wave characteristic of the device scattering parameter in whole wafer, by observing not
With the scattering parameter distribution under working frequency, the scattering parameter under different applied bias is distributed, point of scattering parameter difference component
Cloth, determine that there is the characteristic part of fluctuation or discreteness in the test data；
Step 04：Average value is calculated to the features described above data in all crystal grains, selects the characteristic of surveyed device most
It is golden die close to the crystal grain corresponding to the average value, and tests the scattering ginseng of all device under tests in the golden die
Number, using the device test data of the golden die as fit object, extracts the model parameter in the step 01, establishes
The benchmark model not comprising technological fluctuation of device；
Step 05：Scattering parameter based on the characteristic part and the network converted by scattering parameter are joined
Number, it is configured to one or one group of index that device carries out statistical modeling；
Step 06：Based on the benchmark model in These parameters and step 04, each component value in sub-circuit model is calculated, is gone forward side by side
Sensitivity analysis and sequence of the units value relative to desired value, then choose sensitivity highest element；
Step 07：Based on selected element, to all kinds of ginsengs in the scalable formula of the element in the step 02
Several groups, sensitivity analysis and sequence of the parameter value relative to component value are carried out respectively, chooses sensitivity highest parameter；
Step 08：Based on selected parameter, the statistical property distribution of device is fitted, it is complete for characterizing device so as to obtain
The statistical model of office's fluctuation distribution.
Preferably, in the step 03, the described characteristic part with fluctuation or discreteness is：The test
The ratio between difference and the minimum value of maxima and minima of the data in different crystal grain are more than as the ripple set by modeling requirement
The characteristic part of dynamic threshold value.
Preferably, in the step 05, the index can reflect characteristic of the device under the conditions of radio frequency operation；The spy
Property includes effective capacitance value.
Preferably, in the step 06, sensitivity analysis and sequence of the component value relative to desired value, specific bag are carried out
Include：Near the component value, more all elements are become as index in the step 05 caused by the identical change ratio of component value
The scale of change and sequence.
Preferably, in the step 06, when choosing sensitivity highest element, selected sensitivity highest element
Quantity is to reach the minimum number of elements needed for fit object.
Preferably, the step 07 includes：Respectively to basic parameter group, biasing relevant parameter group, the related ginseng of device size
Several groups carry out sensitivity analysis and sort.
Preferably, the step 07 includes：When choosing sensitivity highest parameter, selected sensitivity highest ginseng
Number quantity is to reach the minimum parameters quantity needed for fit object.
Preferably, in the step 08, used approximating method is to selected parameter addition statistical function and adjusts system
Count functional dependence parameter.
Preferably, the fit procedure in the step 08 specifically includes：First, it is non-for being brought by device size reduction
Index when the minimum device of preferable effect is without applied bias, uses the spirit in basic parameter group selected in the step 07
Sensitivity highest parameter is fitted；Then, whether indicator-specific statistics characteristic when observing device applied bias meets that fitting requires,
As do not met, continue addition and entered using the sensitivity highest parameter in biasing relevant parameter group selected in the step 07
Row fitting；Then, whether the indicator-specific statistics characteristic for observing different scale devices meets that fitting requires, does not meet such as, continues to add
It is fitted using the sensitivity highest parameter in device size relevant parameter group selected in the step 07；In completion
Fit procedure is stated, that is, obtains the statistical model for characterizing the global fluctuation distribution of device.
Preferably, the radio frequency variodenser is MOS varactor.
The statistical modeling method of the radio frequency variodenser of the present invention, for different sub-circuit structures, different scalable formula institutes
The model of composition, the present invention are all suitable for；Also, mapping test data analysis is based on, selection can reflect device key performance
Index, the principal statistical characteristic distribution for the process characteristic both can be effectively held, and ensure that establishing model can react again
The statistical distribution of circuit designers Primary Component performance of concern；Furthermore based on parameter value to component value, component value is to index
The sensitivity analysis of value, and parametric classification is carried out, can fast selecting maximally effective statistical model on the basis of current base model
Parameter；Meanwhile for the independence of different parameters group, extract statistical parameter value successively respectively based on corresponding performance indications, make
Modeling process is quickly clear.
Brief description of the drawings
Fig. 1 is the cross-sectional view of N-type MOS varactor
Fig. 2 is the sub-circuit topological structure schematic diagram of the preferred embodiment of the present invention
Fig. 3 is the schematic flow sheet of the statistical modeling method of the radio frequency variodenser of the preferred embodiment of the present invention
Embodiment
To make present disclosure more clear understandable, below in conjunction with Figure of description, present disclosure is made into one
Walk explanation.Certainly the invention is not limited in the specific embodiment, the general replacement known to those skilled in the art
Cover within the scope of the present invention.
The statistical modeling method of the radio frequency variodenser of the present invention, the radio frequency variodenser are distributed on the crystal grain of wafer, its
Comprise the following steps：
It is determined that the sub-circuit topological structure for organs weight, determine that each circuit element in the sub-circuit uses
Scalability formula, for the modeling of different scale devices, so that it is determined that organs weight needs all model parameters extracted；
Parameter is classified, is divided into basic parameter group, the population of parameters related to applied bias, ginseng related to device size
Several groups, and population of parameters uncorrelated to technological fluctuation；
Based on radio frequency modeling demand, parameter testing is scattered to device, selected section critical size device first is carried out
The mapping tests of scattering parameter, i.e., be all scattered parameter testing to the selected scale device in all crystal grains (die), point
The mapping test datas of scattering parameter are analysed, obtain wave characteristic of the device scattering parameter in whole wafer, by observing not
With the scattering parameter distribution under working frequency, the scattering parameter under different applied bias is distributed, point of scattering parameter difference component
Cloth, determine the characteristic part fluctuated in the test data or dispersion degree is larger；
Average value is calculated to the features described above data in all crystal grains, the characteristic for selecting surveyed device is flat near this
Crystal grain corresponding to average is golden die, and tests the scattering parameter of all device under tests in the golden die, with this
Golden die device test data is fit object, extracts the model parameter in the step 01, establishes device
Benchmark model not comprising technological fluctuation；
Scattering parameter based on the characteristic part and the network parameter converted by scattering parameter, construction are used
One or one group of index of statistical modeling are carried out in device；
Based on These parameters and above-mentioned benchmark model, each component value in sub-circuit model, units value phase of going forward side by side are calculated
Sensitivity analysis and sequence for desired value, then choose sensitivity highest element；
Based on selected element, to all kinds of populations of parameters in the above-mentioned scalable formula of the element, joined respectively
Sensitivity analysis and sequence of the numerical value relative to component value, choose sensitivity highest parameter；
Based on selected parameter, the statistical property distribution of device is fitted, so as to obtain being used for characterizing the global fluctuation of device
The statistical model of distribution.
The statistical modeling method of the radio frequency variodenser of the present invention is made below in conjunction with accompanying drawing 1-3 and specific embodiment further
Describe in detail.It should be noted that accompanying drawing uses very simplified form, using non-accurately ratio, and only to convenient, clear
Reach the purpose for aiding in illustrating the present embodiment clearly.
Referring to Fig. 3, the statistical modeling method of the radio frequency variodenser of the present embodiment, comprises the following steps：
Step 01：It is determined that the sub-circuit topological structure for organs weight, determines that each circuit element makes in sub-circuit
Scalability formula, for the modeling of different scale devices, so that it is determined that organs weight needs all models extracted
Parameter；
Specifically, referring to Fig. 1, for a kind of cross-sectional view of N-type MOS varactor, formed in P type substrate 1
There is N-type trap 2, the grid oxide layer 4 in N-type trap 2, the grid 5 being covered on grid oxide layer 4, and the N-type positioned at the both sides of grid 5
Source/drain region 3 in trap 2.Modeling of the industry for variodenser at present uses the form of sub-circuit, the sub-circuit for modeling mostly
Topographical form is various.The sub-circuit topological structure that the present embodiment is taken is as shown in Figure 2.The topological circuit structure includes 10 altogether
Individual circuit element, it is Lg, Rg, Cs, Cf, Rch, Rsd, Rwell, Csub1, Rsub and Csub2 respectively.
Specifically, in order to meet the needs that are emulated to different scale devices, it is necessary to establish the scalable model of device,
Scalable formula is established to each circuit element to be fitted the test data of different scale devices.Such as Lg=a1*Wa2*
La3, a1, a2, a3 be in the formula it needs to be determined that model parameter.These formula can be by modeling personnel according to device physicses work
Make mechanism and test data voluntarily determines.Acceptor circuit structure does not limit the method for the invention, also not each in by sub-circuit
The form limitation of the fitting formula of circuit element.The model parameter extracted is all model parameters.
Step 02：Parameter is classified, is divided into basic parameter group, the population of parameters related to applied bias, with device chi
Very little relevant parameter group, and the population of parameters unrelated with technological fluctuation；Here, the population of parameters unrelated with technological fluctuation can be temperature
Property dependent parameter group.
Step 03：Based on radio frequency modeling demand, parameter testing, first selected section critical size device are scattered to device
Part is scattered the mapping tests of parameter, i.e., is all scattered parameter testing to the selected scale device in all crystal grains, point
The mapping test datas of scattering parameter are analysed, obtain wave characteristic of the device scattering parameter in whole wafer, by observing not
With the scattering parameter distribution under working frequency, the scattering parameter under different applied bias is distributed, point of scattering parameter difference component
Cloth, determine that there is the characteristic part of fluctuation or discreteness in test data, selection fluctuation or discrete journey in the embodiment
Spend larger characteristic part；Specifically identified degree of fluctuation can flexibly be determined by the distribution situation of actual test data,
Degree of fluctuation in the present embodiment is defined as the difference and minimum of maxima and minima of the test data in different crystal grain
The ratio between value is more than fluctuation threshold set in advance.The threshold value can be flexibly true by the actual distribution feature of modeling requirement and test data
Fixed, the present embodiment is taken as 5%.Here, crystal grain industry is commonly referred to as die.
Step 04：Average value or median are calculated to the features described above data in all crystal grains, select the spy of surveyed device
It is golden die that data, which are levied, near the crystal grain corresponding to the average value or median, and tests in the golden die and own
The scattering parameter of device under test, using the device test data of the golden die as fit object, extract in the step 01
Model parameter, establish the benchmark model not comprising technological fluctuation of device；Here, golden die are properly termed as optimal crystal grain.
Step 05：The S parameter of feature based data division and the network parameter converted by S parameter, are configured to
Device carries out one or one group of index of statistical modeling；
Specifically, here, it is determined that the evaluation index for global statistics modeling.The index should reflect the main of test data
Statistical distribution characteristic, and key property of the device in the range of radio frequency operation such as effective capacitance value can be reflected, can be one
Individual or one group of index.It is distributed by observing the S parameter under different operating frequency, the S parameter distribution under different applied bias, S ginsengs
The distribution of the different components of number, determines to fluctuate larger data division in device test data.And based on S parameter and by S parameter
Other network parameters such as impedance parameter (Z parameter), admittance parameter (Y parameter) etc. converted, is configured to device and is united
Count one or one group of index of modeling.The present embodiment uses
As the index for judging variodenser radio frequency statistical property.Wherein, Z parameter is impedance parameter, for two-port network,
Four components of Z parameter are plural number, respectively Z11, Z12, Z21, Z22；Imag function representations take imaginary part, that is, take (Z11-
Z12 imaginary part)；ω is angular frequency, and ω=2 π * f, f are frequency；ω be able to can also be selected with the key job frequency of selector
One group of ω is selected, seeks the weighted average of F under different ω as index.Preferably, the present embodiment calculates this using the π * 5E9 of ω=2 and referred to
Mark.
Step 06：Based on the benchmark model in These parameters and step 04, each component value in sub-circuit model is calculated, is gone forward side by side
Sensitivity analysis and sequence of the units value relative to desired value, then choose sensitivity highest element；
Specifically, sensitivity analysis and sequence of the component value relative to desired value are carried out, including：Near component value, than
More all elements are as the scale of index change and calculating of sorting in the step 05 caused by the identical change ratio of component value
Sensitivity and sequence of each component value to the index, choose sensitivity highest element and carry out statistical modeling.When selection sensitivity
During highest element, selected sensitivity highest number of elements is to reach the minimum number of elements needed for fit object,
It is that selected element is that can meet fitting demand, the minimal number of element chosen successively by sensitivity height.For example, according to reality
Border fit solution, fit object can be reached if only needing an element, just only with one, can not such as meet to add successively sensitive
The second high, the 3rd high element is spent, until disclosure satisfy that fitting demand or new addition element can not play a significant role.
It should be noted that there are many kinds for the computational methods of sensitivity.The strict calculating of sensitivity should take statistics to refer to
Scale value treats analysis element and carries out derivation respectively, but immediate derivation is more tired in such a more complicated sub-circuit structure
It is difficult.A kind of numerical method is taken in the present embodiment, i.e., based on benchmark model, calculates all component values, element to be analyzed is in the base
Change certain proportion on the basis of quasi-mode offset a0, such as i.e. two of change are worth component value up and down at a given proportion on the basis of a0,
A1=a0 (1- Δs), a2=a0 (1+ Δs) are can be expressed as, here Δ=5%, remaining component value is constant, counting statistics index F0's
Change ratio (two of change are worth component value up and down at a given proportion on the basis of a0), that is, calculating F1, (component value is counted when being a1
The statistical indicator of calculation), F2 (statistical indicator that component value calculates when being a2).The calculating of sensitivity is exactly the change ratio of statistical indicator
The change ratio of example divided by component value to be analyzed；Sensitivity α can calculate according to following formulas：
α=︱ (F1-F2)/F0 ︱/︱ (a 1-a2)/a0 ︱.
By calculating, the present embodiment chooses Cs, and statistical model modeling is carried out with two elements of Rch.
Step 07：Based on selected element, to all kinds of populations of parameters in the scalable formula of element in step 02, difference
Sensitivity analysis and sequence of the parameter value relative to component value are carried out, chooses sensitivity highest parameter；
Specifically, the parameter for statistical modeling is determined, respectively to basic parameter group, biasing relevant parameter group, Yi Jiqi
Part size relevant parameter group is analyzed and sorted, and selects sensitivity highest parameter；When choosing sensitivity highest parameter,
Selected sensitivity highest number of parameters is to reach the minimum parameters quantity needed for fit object, that is to say selected ginseng
Number is that can meet fitting demand, the minimal number of parameter chosen successively by sensitivity height.For example, it is fitted feelings according to actual
Condition, fit object can be reached if only needing a parameter, just only with one, can not such as meet to add sensitivity second successively
Height, the 3rd high element, until disclosure satisfy that fitting demand or new addition parameter can not play a significant role.
For the element of selection, influence of the parameter in the scalable formula of the element to element value changes is analyzed respectively,
So that it is determined that the parameter of statistical disposition should be done.In the present embodiment below equation is used for Cs and Rch：
In above-mentioned formula, w, l, area, pj are the physical dimension of device, and Vgs is the operating voltage of device, and T is device
Temperature during work, remaining variables are model parameter, and model parameter can be divided into several groups, and the first kind is basic parameter such as Cmp Cml
Cma dCp dCl dCa, Rch0 Rbeta etc., the second class is with biasing relevant parameter such as dvgs0 vgn0 dVch KaccDeng the 3rd
It with device size relevant parameter such as dl, the 4th class is the parameter unrelated with technological fluctuation that class, which is, such as temperature characterisitic relevant parameter
Such as dvgs_temp, vgn_temp etc..
It should be noted that Parameter Sensitivity Analysis can equally take the numerical analysis method in above-mentioned steps 06.
Step 08：Based on selected parameter, the statistical property distribution of device is fitted, it is complete for characterizing device so as to obtain
The statistical model of office's fluctuation distribution.
Specifically, used approximating method is to add statistical function to selected parameter and adjust statistical function correlation to join
Number, here, adjusts each parameter and finds out their 3 σ values, determine their fluctuation pattern.Use P=P0+P_3 σ * agauss
(0,1,3) original parameter P0 is replaced, wherein, P_3 σ represent to need the parameter for adjusting fitting, and agauss is normal distyribution function, is also cried
Gauss of distribution function.0 represents mean μ represents 3 times of σ of standard deviation 3 as 1 as 0,1 and 3.P_3 σ numerical value is adjusted, model is imitated
The true statistical distribution for making model emulation data levels off to the statistical distribution of the test data in institute's step 03；
Fit procedure specifically includes：
First, (it is usually institute for reducing the minimum device of the non-ideal effects brought by size in device device under test
Design test structure in size maximum device, hereinafter referred large-size device) without applied bias when index, using in step 07
Sensitivity highest parameter in selected basic parameter group is fitted；Then, index during device applied bias is observed
Whether statistical property meets that fitting requires, does not meet such as, continues addition and uses biasing relevant parameter group selected in step 07
In sensitivity highest parameter be fitted；Then, whether the indicator-specific statistics characteristic for observing different scale devices meets to be fitted
It is required that as do not met, continue addition and use the sensitivity highest in device size relevant parameter group selected in step 07
Parameter is fitted；Above-mentioned fit procedure is completed, that is, obtains the statistical model for characterizing the global fluctuation distribution of device.
For example, carry out statistical model fitting, it is first determined sensitivity sequence of the basic parameter group to component value, select sensitive
Spend highest parameter and carry out models fitting.Here, dC is selected respectivelylStatistical property fitting is carried out to large-size device with Rbeta.
Then observe whether the model can be fitted data of the device in applied bias, as that can not be fitted, it is related to be continuing with bias
Population of parameters carries out models fitting, equally enters line sensitivity sequence, and the present embodiment chooses vgn0 and KaccIt is fitted.Observe afterwards
Whether the statistical property of different scale devices is fitted, and the fitting of size relevant parameter, this implementation should be continuing with by such as failing fitting
Example is fitted using dl.The present embodiment does not consider statistical property difference during different temperatures characteristic.By said process, the system
Meter model meets the statistical property distribution of the lower device of the different biasings of different sizes.
In summary, the statistical modeling method of radio frequency variodenser of the invention, for different sub-circuit structures, difference can stretch
The model that contracting formula is formed, the present invention are all suitable for；Also, mapping test data analysis is based on, selection can reflect that device closes
The index of key energy, both can effectively hold the principal statistical characteristic distribution for the process characteristic, ensure to establish model energy again
Enough statistical distributions for having reacted circuit designers Primary Component performance of concern；Furthermore based on parameter value to component value, element
Be worth sensitivity analysis to desired value, and carry out parametric classification, can fast selecting it is maximally effective on the basis of current base model
Statistical model parameter；Meanwhile for the independence of different parameters group, statistics ginseng is extracted successively respectively based on corresponding performance indications
Numerical value, make modeling process quickly clear.
Although the present invention is disclosed as above with preferred embodiment, the right embodiment illustrated only for the purposes of explanation and
, the present invention is not limited to, if those skilled in the art can make without departing from the spirit and scope of the present invention
Dry change and retouching, the protection domain that the present invention is advocated should be to be defined described in claims.
Claims (10)
1. a kind of statistical modeling method of radio frequency variodenser, the radio frequency variodenser are distributed on the crystal grain of wafer, its feature exists
In comprising the following steps：
Step 01：It is determined that the sub-circuit topological structure for organs weight, determines that each circuit element makes in the sub-circuit
Scalability formula, for the modeling of different scale devices, so that it is determined that organs weight needs all models extracted
Parameter；
Step 02：Parameter is classified, is divided into basic parameter group, the population of parameters related to applied bias, with device size phase
Related parameter group, and the population of parameters unrelated with technological fluctuation；
Step 03：Based on radio frequency modeling demand, parameter testing is scattered to device, selected section critical size device first enters
The mapping tests of row scattering parameter, i.e., be all scattered parameter testing to the selected scale device in all crystal grains, and analysis dissipates
The mapping test datas of parameter are penetrated, wave characteristic of the device scattering parameter in whole wafer are obtained, by observing different works
Scattering parameter under working frequency is distributed, the scattering parameter distribution under different applied bias, the distribution of scattering parameter difference component, really
There is the characteristic part of fluctuation or discreteness in the fixed test data；
Step 04：Average value is calculated to the features described above data in all crystal grains, select the characteristic of surveyed device near
Crystal grain corresponding to the average value is optimal crystal grain golden die, and is tested all to be measured in the optimal crystal grain golden die
The scattering parameter of device, using the device test data of the optimal crystal grain golden die as fit object, extract the step 01
In the model parameter, establish the benchmark model not comprising technological fluctuation of device；
Step 05：Scattering parameter based on the characteristic part and the network parameter converted by scattering parameter, structure
Make one or the one group of index that statistical modeling is carried out for device；
Step 06：Based on the benchmark model in These parameters and step 04, each component value in sub-circuit model is calculated, and carry out member
Sensitivity analysis and sequence of the part value relative to desired value, then choose sensitivity highest element；
Step 07：Based on selected element, to all kinds of populations of parameters in the scalable formula of the element in the step 02,
Sensitivity analysis and sequence of the parameter value relative to component value are carried out respectively, choose sensitivity highest parameter；
Step 08：Based on selected parameter, the statistical property distribution of device is fitted, so as to obtain being used for characterizing device overall situation ripple
The statistical model of dynamic distribution.
2. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that in the step 03, institute
The characteristic part with fluctuation or discreteness stated is：Maximum of the test data in different crystal grain and minimum
The ratio between difference and the minimum value of value are more than as the characteristic part of the fluctuation threshold set by modeling requirement.
3. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that in the step 05, institute
Characteristic of the device under the conditions of radio frequency operation can be reflected by stating index；The characteristic includes effective capacitance value.
4. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that in the step 06, enter
Sensitivity analysis and sequence of the units value relative to desired value, are specifically included：Near the component value, more all elements
The scale changed as index in the step 05 caused by the identical change ratio of component value and sequence.
5. the statistical modeling method of radio frequency variodenser according to claim 4, it is characterised in that in the step 06, when
When choosing sensitivity highest element, selected sensitivity highest number of elements is to reach the minimum member needed for fit object
Number of packages amount.
6. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that the step 07 includes：
Sensitivity analysis is carried out to basic parameter group, biasing relevant parameter group, device size relevant parameter group respectively and sorted.
7. the statistical modeling method of radio frequency variodenser according to claim 6, it is characterised in that the step 07 includes：
When choosing sensitivity highest parameter, selected sensitivity highest number of parameters is to reach minimum needed for fit object
Number of parameters.
8. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that in the step 08, institute
The approximating method used is adds statistical function to selected parameter and adjusts statistical function relevant parameter.
9. the statistical modeling method of radio frequency variodenser according to claim 8, it is characterised in that the plan in the step 08
Conjunction process specifically includes：When first, for reducing the minimum device of the non-ideal effects brought without applied bias by device size
Index, be fitted using the sensitivity highest parameter in basic parameter group selected in the step 07；Then, see
Whether indicator-specific statistics characteristic when examining device applied bias meets that fitting requires, does not meet such as, continues addition and uses the step
Sensitivity highest parameter in 07 in selected biasing relevant parameter group is fitted；Then, different scale devices are observed
Indicator-specific statistics characteristic whether meet that fitting requires, do not meet such as, continue addition and use selected device in the step 07
Sensitivity highest parameter in size relevant parameter group is fitted；Above-mentioned fit procedure is completed, that is, obtains being used for tokenizer
The statistical model of the global fluctuation distribution of part.
10. the statistical modeling method of radio frequency variodenser according to claim 1, it is characterised in that the radio frequency variodenser
For MOS varactor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510111705.5A CN104679960B (en) | 2015-03-13 | 2015-03-13 | A kind of statistical modeling method of radio frequency variodenser |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510111705.5A CN104679960B (en) | 2015-03-13 | 2015-03-13 | A kind of statistical modeling method of radio frequency variodenser |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104679960A CN104679960A (en) | 2015-06-03 |
| CN104679960B true CN104679960B (en) | 2018-04-03 |
Family
ID=53314999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510111705.5A Active CN104679960B (en) | 2015-03-13 | 2015-03-13 | A kind of statistical modeling method of radio frequency variodenser |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104679960B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105760604B (en) * | 2016-02-19 | 2019-01-18 | 上海集成电路研发中心有限公司 | The modeling method of statistical model based on domain kindred effect |
| CN106249120B (en) * | 2016-08-02 | 2018-12-18 | 上海集成电路研发中心有限公司 | The selection method of Center Wafer for model data test |
| CN108169248B (en) * | 2016-12-07 | 2020-02-18 | 成都理工大学 | Self-adaptive S parameter matching method for detecting liquid dangerous goods |
| CN107480331B (en) * | 2017-07-07 | 2020-10-16 | 中国科学院微电子研究所 | Modeling method and device of semiconductor device statistical model |
| CN109657390A (en) * | 2018-12-28 | 2019-04-19 | 中国电子科技集团公司第二十九研究所 | A kind of technique IP statistical modeling method in radio frequency Integrated manufacture |
| CN111859844A (en) * | 2019-04-26 | 2020-10-30 | 长鑫存储技术有限公司 | Feature chip selection method and device model establishment method in modeling process |
| CN110222087B (en) * | 2019-05-15 | 2023-10-17 | 平安科技（深圳）有限公司 | Feature extraction method, device and computer readable storage medium |
| CN114841099B (en) * | 2022-07-04 | 2022-10-11 | 浙江铖昌科技股份有限公司 | Method, device, equipment and system for constructing characterization model |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001244452A (en) * | 2000-02-28 | 2001-09-07 | Japan Radio Co Ltd | Method and apparatus for extracting high-frequency semiconductor device model parameter, and recording medium |
| CN101196936A (en) * | 2006-12-05 | 2008-06-11 | 上海华虹Nec电子有限公司 | Fast modeling method of MOS transistor electricity statistical model |
| CN101655882A (en) * | 2009-07-24 | 2010-02-24 | 上海宏力半导体制造有限公司 | Modelling method based on worst condition of statistic model |
| CN102117352A (en) * | 2010-01-05 | 2011-07-06 | 上海华虹Nec电子有限公司 | Method for simulating radio frequency metal oxide semiconductor (MOS) varactor |
-
2015
- 2015-03-13 CN CN201510111705.5A patent/CN104679960B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001244452A (en) * | 2000-02-28 | 2001-09-07 | Japan Radio Co Ltd | Method and apparatus for extracting high-frequency semiconductor device model parameter, and recording medium |
| CN101196936A (en) * | 2006-12-05 | 2008-06-11 | 上海华虹Nec电子有限公司 | Fast modeling method of MOS transistor electricity statistical model |
| CN101655882A (en) * | 2009-07-24 | 2010-02-24 | 上海宏力半导体制造有限公司 | Modelling method based on worst condition of statistic model |
| CN102117352A (en) * | 2010-01-05 | 2011-07-06 | 上海华虹Nec电子有限公司 | Method for simulating radio frequency metal oxide semiconductor (MOS) varactor |
Non-Patent Citations (2)
| Title |
|---|
| 基于BSIM3模型的毫米波MOS变容管建模;夏立诚 等;《微电子学》;20071231;第37卷(第6期);第833-837页 * |
| 射频集成（RFIC）压控振荡器（VCO）的设计与研究;易新敏;《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》;20061215(第12期);第I135-343页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104679960A (en) | 2015-06-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104679960B (en) | A kind of statistical modeling method of radio frequency variodenser | |
| TWI719804B (en) | Method of optical metrology, computer program product, metrology module, target design file, landscape and metrology measurements of targets | |
| US10726169B2 (en) | Target and process sensitivity analysis to requirements | |
| US8997028B2 (en) | Methods and system for analysis and management of parametric yield | |
| CN103226627B (en) | Method and device for simulating chip surface morphology | |
| US7739065B1 (en) | Inspection plan optimization based on layout attributes and process variance | |
| US20140033159A1 (en) | Method of Optimizing Design for Manufacturing (DFM) | |
| US20080262769A1 (en) | Using multivariate health metrics to determine market segment and testing requirements | |
| TWI672753B (en) | System and method for field-by-field overlay process control using measured and estimated field parameters | |
| CN108089412A (en) | Photoetching alignment precision measures the appraisal procedure of accuracy | |
| TW202134601A (en) | Accuracy improvements in optical metrology | |
| CN108352336A (en) | Determine one or more characteristics of pattern of interest on sample | |
| CN106158679A (en) | In conjunction with wafer entity measuring with digital resolution to improve semiconductor element manufacturing method thereof | |
| Stratigopoulos et al. | Evaluation of analog/RF test measurements at the design stage | |
| CN106407490A (en) | System and method for discovering unknown problematic patterns in chip design layout | |
| CN104779144B (en) | Improve the design feature analytic approach of the micro-photographing process of manufacture semiconductor subassembly | |
| KR20200096992A (en) | Semiconductor measurement and defect classification using electron microscope | |
| CN100498548C (en) | Method for predicting function of integrated circuit fragment photo-etched on chip | |
| CN104778181B (en) | A kind of method and its equipment measuring spectrum and library Spectral matching | |
| CN109426655A (en) | Data analysing method, device, electronic equipment and computer readable storage medium | |
| Chan et al. | Design dependent process monitoring for back-end manufacturing cost reduction | |
| Cilici et al. | Assisted test design for non-intrusive machine learning indirect test of millimeter-wave circuits | |
| CN107924850A (en) | Self orientation is measured and pattern classification | |
| CN101556930B (en) | Technique for evaluating a fabrication of a semiconductor component and wafer | |
| CN108693714A (en) | OPC data collection method and OPC data collection device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |