SPCClass 1Class 1IntroductionDavid Parent1Agenda• Goals of the course•QualityQuality• Control ChartsWh th d–Why they are used–Example2Shewhart and DemingThe long range contribution of statistics depends notThe long range contribution of statistics depends not so much upon getting a lot of highly trained statisticians intoindustry as it does in creating a statistically minded generation ofphysicists, chemists, engineers, and other who will in any way have a hand in developing and directing the production processes of tomorrow.3Goals of this course• Students should be able to–Speak the language of SPCSpeak the language of SPC– Incorporate its principles into designs–Know how to use an“SPC design flow”Know how to use an SPC design flow• develop a control chart or an acceptance procedure4More specifically• Develop R, s, and p control charts•know when to use eachknow when to use each• Use a control chart on data gathered in class to improve the 129 or 167 process lineto improve the 129 or 167 process line• Use control charts to make decisions or di ireduce inspection costs5Quality• quality of specification•quality of conformancequality of conformance• degree of excellence?li (5 h )• quality (5 ohms)6Control chart fundamentals• Measured quality of manufactured products is always subject to a certain amount of variation as a result of chance• Some “system of chance causes” is inherent in any il h f d i di iparticular scheme of production and inspection• Variation with in a stable pattern is inevitable.• The reasons for variation outside this stable pattern may be discovered and corrected.7Cooperation• Design engineers and manufacturing engineers some times do not talk.g– Specifications can be set tighter than the system can produce.p– Specification can be pulled out of this air–SPC can show you what is possible and if what ypis possible will work! It come from hard data and not opinion!8Types of charts•Xbar, R•Xbar s• fraction rejected (p)• nonconforming (c)g( )•Variable–continuous length, width, current, powerg, , ,p• Attribute–counts or articles or events9Language• Defect vs defective•nonconformance: not meeting onenonconformance: not meeting one specification•percent rejected rather than percent•percent rejected rather than percent defective10Control ChartsControl ChartsEE/MatE167David W. ParentDavid W. Parent11Control Chart Design FlowI. Make decisions to prepare for control chart constructionA. Possible ObjectivesB. Choose the variableC. Decide on basis of subgroupD. Choose size and frequency of subgroupsE. Set the data collection formF. Determine method of measurement12Control Chart Design FlowII. Start the control chartsA.Make chart and record dataA.Make chart and record dataB. Calculate Xbar and R for each subgroupCPlot Xbar and RC.Plot Xbar and R13Control Chart Design FlowIII. Determine trial control limitsA.Decide on how many subgroups are requiredA.Decide on how many subgroups are required to calculate the limitsB. Calculate trial control limitsC. Plot the central lines and limits14Control Chart Design FlowIV. Draw preliminary conclusionsA.Decide whether in control or outA.Decide whether in control or outB. InterpretationCDiscuss relationship between out of controlC.Discuss relationship between out of control processes and specification limitsD.Suggest actionsD.Suggest actions15Control Chart Design FlowV. Continue to use the chartsA.Revise central lines and limitsA.Revise central lines and limitsB. Sort lotsCDecide on an action about a processC.Decide on an action about a processD. Acceptance inspectionEVerify specificationsE.Verify specifications16I-A Objectives• Gather data–to verify specificationsto verify specifications– verify production procedures–verify inspection proceduresverify inspection procedures– justify current decisionsjustify whether to hunt for problems–justify whether to hunt for problems– Learn control charts17I-B Variable• Gather data that can–be measured in numbersbe measured in numbers– that can be used to save money or improve processp18I-C sub group• Choose subgroups that–are as homogenous as possible if you are tryingare as homogenous as possible if you are trying to detect shifts in the process mean–are not homogeneous if acceptance testinggpg19I-D Size of subgroup• 4 is thought to be ideal•5 is easy to calculate5 is easy to calculate– gets all area of a wafertry to minimize variation in subgroup–try to minimize variation in subgroup• 2 or 3 if it is expensive• 10-20 if we need a large sensitivity of shifts in the mean and standard deviation20Forms and method ofForms and method of measurements•Forms–We will try to develop standard excell sheets soWe will try to develop standard excell sheets so groups can share data•MeasurementMeasurement– We have detailed instructions on how to use equipmentequ p e t21Example• EE129 process•Assume we are making simple currentAssume we are making simple current mirrors.22I-A Objectives• What are the possible objectives?•Or what are our main problems inOr what are our main problems in EE/MatE129?Currently we just make transistors to learn–Currently we just make transistors to learn about processing, we don’t make them for a given applicationgpp– We want to change this so EE students can design circuits that are made in 129.23gWhat do we need for a designWhat do we need for a design environment?• A process that is in control•An idea of what specification limits theAn idea of what specification limits the circuits designers can expect24More possible objectivesMore possible objectives (Current Decisions)• Currently in 129 there is a lot of time spent measuring oxide thicknesses for etching, lab often goes over this time or students have to come backgoes over this time or students have to come back to do it. –We want to reduce time spent inspecting! ppg• Currently in the 129 process there is 8 microns of overlap to prevent device failures to to alignment–This caused the S/D area to be large• This causes device to be big, which limits how many devices we can put a on chip25• More leakage– We need to verify this design rule!More possible objectives• Since we have not been doing control charts, we do not know if we have a ,problems, so lets not hunt for any.–If we find one, we find one.If we find one, we find one.• We need to learn control