Title: Integrated Circuits Cycle Time at Test Operation Estimation for Production Planning using Computer Simulation By Jitrayut Junnapart
1Integrated Circuits Cycle Time at Test Operation
Estimationfor Production Planningusing Computer
SimulationByJitrayut Junnapart
2Introduction
3Introduction(1) Assembly Process
- General Process of IC manufacturing
4Introduction(2) Electrical Test Process
- All products must be 100 tested before deliver
to customers - Electrical Test process is shown below
5Introduction(3) Electrical Test Time
- Test Time depends on
- Test Rate at main flow (in Units Per Hour)
- First Pass Yield (0 to 1 for 0 to 100)
- Test Rate at rescreen flow (in UPH)
To P.15
6Introduction(4) An Example
- Suppose one production lot of
- 3,000 units
- Main Flow Test Rate 1,000 UPH (Units Per
Hour) - Rescreen Flow Test Rate 300 UPH
- First Pass Yield 90
Main Flow Test Time 3,000 / 1,000
3 hours Rescreen Flow Test Time 3,000 (1 -
0.9) / 300 1 hour Total Test Time 3
hours 1 hour 4 hours
7Introduction(5) Why interesting?
- Electrical test time is very important because it
dominates overall IC manufacturing cycle time - There are methods to calculate cycle time
- Use average values (of UPHs, FPY)
- Apply simulation method
- Others
8Literature Review
9Literature Review(1)
- This project applies Monte Carlo Simulation based
on historical data of - Main flow test rate
- Rescreen flow test rate
- First pass yield
- to simulate for total cycle time of one
production lot
10Literature Review(2)
- Monte Carlo simulation uses random number to
pick one value of data as a representative for
further calculation
11Literature Review(3) Example
- Suppose we want to simulate main flow test time
- And partial of historical data of main flow test
time (in UPH) looks like
12Literature Review(4) Step 1
- Rearrange the numbers into frequency and
probability table
13Literature Review(5) Step 2
- Construct cumulative probability table from
probability table
14Literature Review(6) Step 3
- Set up probability intervals for ranges of UPH
15Literature Review(7) Step 4
- Generate a random number and map onto the
appropriate interval to find a simulated value of
main flow UPH
Random No. generated is 0.5934
Simulated UPH 160
16Literature Review(8) Step 5
- Repeat step 1-4 but for
- First Pass Yield and
- Rescreen Flow UPH
- Use equation in slide 5 to calculate Simulated
Total Test Time
17Literature Review(9)
- BUT real test rate (UPH) data is continuous
number (appendix C,p79) - We need to group them first into ranges
- Then count number of data that are in each range
- And find the center value of the group as a
representative number of the group
18Literature Review(10)
- Number of group can be chosen
- Groups shall cover every data
- See appendix D, p.86 for detail
19Project Development
20Project Development (1)
- Problem Definition
- Electrical test time is the main dominator of
overall production cycle time in back-end IC
manufacturing - Cycle time affects shipping commitment to
customers - Using average test time is not effective because
the numbers vary over wide range
Production control people wants a tool to
forecast electrical test time of the IC
21Project Development (2)
- Objectives
- To develop computer software that processes
historical test rate data (in Units Per Hour -
UPH) at main flow and rescreen flow, and first
pass yield (retest rate-related) from production
and simulate for probability to finish testing a
production lot with various allocated time. - The computer software must be separated into main
program and auxiliary data files. The main
program will work with various environments
without modification. - The developed simulation program code must work
on both Personal Computer (PC) and larger
computer running on UNIX operating system.
22Project Development (3)
- Scope and limitations of the project
- It is considered that main flow test rate,
rescreen flow test rate, and first pass yield
being input of this project are available
beforehand. They are stored in separated text
files which, in fact, easy to develop the system
further. - This project gives calculation output as a text
file for easy to use in the future. Example of
this is putting into Hypertext Markup Language
(HTML) and post on facilitys Intranet. - The software in this project will calculate for
success probability to finish electrical test
process based on a certain size of production
lot. User can later use linear relationship to
calculate for other size.
23Project Development (4)
- Project Time Line
- The project consists of 9 major tasks
- Total time taken is 7.5 months
24Project Development (5)
- Feasibility Study
- Technical Feasible
- The champion has knowledge of simulation and
programming language - Economic Feasible
- The project does not require special expenses
- Operational Feasible
- The project does not interfere the normal work
practice of concerning people - Legal Feasible
- Software used in development is legal and all
data is legally used
25Project Implementation
26Project Implementation (1)
- Output must be easy for user to understand
- Test time depends on lot size, and it is not
feasible to provide all lot sizes output - User should have control over some essential
simulation parameters - Change in source code is undesirable
1,2
3
4
Need modifications
27Project Implementation (2)
- To solve concerns 1 and 2
To concerns
- An allocated time is set
- The simulated test time for a certain lot size is
calculated and compared to a allocated time. If
test time is shorter than allocated time, the
result is possible. This 'success number' is
counted. - Repeat step 2 for many times and calculate
success chance from success times and total
simulation times and report this chance to user. - Increase the allocated time by delta defined
- Repeat step 2 to 4 for certain rounds defined
28Project Implementation (3)
To concerns
- Store simulation parameters in file para.txt
- Number of types of IC to perform simulation
- Number of divisions to group historical data
- Lot size that simulation is based on
- Number of repetitive simulation for one allowable
time - Start point of allowable time
- Increment of allowable time
- Steps of increment of allowable time
29Project Implementation (4)
To concerns
- All input data are stored in separated files
- Historical files of Main flow UPH (.mnf)
- Historical data of Rescreen flow UPH (.rsf)
- Historical data of First pass yield (.rrf)
- List of IC types file (progname.lst)
30Project Implementation (5) Files
- Main flow UPH file
- icname.mnf
- Rescreen flow UPH file
- icname.rsf
- First pass yield file
- icname.rrf
- List of icname file progname.txt
31Project Implementation (6)
Open file progname.lst and para.txt that
stores test program names to simulate and read
all the names and parameters
Processed for every test program?
End
Yes
No
1.Open file that stores UPH Main flow, and
read all main flow UPH for next program 2. Count
number of data 3. Divide into divisions 4.
Calculate for accumulated probability 5. Repeat
step 1-4 but for rescreen rate and UPH of
rescreen flow
1.Simulate total time used to test the production
lot 2.Count for number of success that the time
used is less than allocated time 3. Repeat 1
and 2 and calculate success probability 4.Record
result in output file 5.Increase allocated time
Simulation Method
Allocated time gt limit?
No
Yes
32Project Implementation (7)
- Programming
- Coding is done in ANSI C language
- Length of the program 900 lines
- Special statements and commands are avoided for
the program can run both on PC and UNIX based
computer system
33Simulation Result
34Simulation Result (1)
1) Screen output
35Simulation Result (2)
2) File output (result.txt)
36Simulation Result (3)
Compiled program
Control parameters input Printed out for
monitoring
Screen output
37Simulation Result (4)
Output file Generated
38Simulation Result (5)
bkkeispemg34 vi result.txt
39Project Evaluation
40Project Evaluation (1)
- Verification of Objectives
- Output is easy for non-engineering people to
understand ? Yes - Control parameters used in simulation can be
input to program through external file ? Yes - Various simulations can be performed without any
changes to main program ? Yes - The source program can be compiled to work on
both PC and UNIX based computer ? Yes
41Project Evaluation (2)
- The simulated result is verified back to
historical data
- The result is 3-5 deviation
- This is very good for simulation (probabilistic
work)
42Conclusions
- Simulation technique can generate result that
gives adequate accuracy (3-5 error in this
work). The result can be transform in the way
that it is easy to understand by end user. - The planner (user) can assign allowable time to
test production lot based on his/her confidence
level that the lot will finish in time - In this particular work, essential parameters in
simulation can be changed easily in control files
without modification to the main program
43Random Number Generation
- Confirm the even distribution by JMP
1,000 numbers
10,000 numbers