Professional Documents
Culture Documents
page
Intel Confidential
Pre-Requisites
Prior to starting Module 2 of this course, you will need to have completed the Web based
training (WBT) for the XCCB website.
It is assumed you know how to initiate a PWP from the XCCB website including filling
out the meta data as well as PWP content.
If you have not done so already, please pause this course and complete the WBT for this
course:
xCCB V2 Overview for Authors ( My Learning Course ID: 10001910, Version: 1.0)
Intel Confidential
Designing an IQ Experiment
You need to write a PWP to bring up a new IHS link in the VF.
B IHS 05 G IHS 03
IHS Example:
Use separate rows for multiple chart types (lot/run level data).
Use (run-to-run) suffix for x-bar charts.
Use (within-run) suffix for s charts.
Include all that are applicable:
– Stability, capability, yield/yield loss assessments
– Reliability and PUPH (pure units per hour) studies.
Intel Confidential DASM – Data Analytics and Statistical Methods Group 13
Section 10 – IHS Link Qual Example
There are several other possible QCs in IHS such as Offset in Y Direction, Flux weight, Lid
Tilt, etc.
Measurement method is how you will directly measure the quality characteristic.
Ideal target or spec limits represents ideal state of each quality characteristic.
• Ideal target for yield is 100%. Assist rate or defect rate has ideal target is 0%.
• Use Ideal target “as high as possible” for a QC that is ideally infinitely large.
Include present mean and standard deviation for each quality characteristic.
Note the source of reference data below the table.
Intel Confidential DASM – Data Analytics and Statistical Methods Group 16
Section 10 – IHS Link Qual Example
Stable
– Should not include any process shifts, trends, extreme outliers, or any other special cause variations.
The SOS script can be used to detect: An acceptable reference dataset can:
• Sample size issues • Be slightly off target
• Lack of stability in the reference data set • Have %OOC > 0%
• False replicates/imputed values • Do not exclude OOC’s if they are
unexplainable
• Show time trends if they are a normal
part of the process
If after several tool quals, the new tool always seems to be much better.
Redefine reference dataset or reference tool to the best in the fleet.
• Does this meet the criteria of a good reference dataset? (Target = 14)
Intel Confidential DASM – Data Analytics and Statistical Methods Group 24
Explanation: Singulation Reference Dataset 3
• During the singulation process, the blade wears as it cuts.
– Blades need to be replaced periodically.
• If the blade naturally wears on its sides instead of its edge, the size of
the units will grow as more product is processed.
– Blades were replaced twice during the data collection (identified on chart)
Note: target = 0
Open the reference data set found in JMP under Process Matching (IQ) Data sets.
Using your knowledge of JMP and DOE1 learnings, complete the empty cells in the table
on the next page.
Enter your results with three decimal places.
First
Second
Third
List the noise factor and the range of possible values for each noise factor.
• Example - substrate supplier as a noise factor has Shinko, Ibiden, and UMTC.
For a noise factor that have unlimited ranges, enter “multiple” to document that multiple
levels are possible.
Intel Confidential DASM – Data Analytics and Statistical Methods Group 34
Section 11 Noise Factors – IHS Example
For scripted factors, determine the number of levels. 2-4 levels are recommended.
. Intel Confidential DASM – Data Analytics and Statistical Methods Group 35
Section 11 Noise Factors - Completed
Accept criteria
Experimental flow
Since IQ experiments are standard designs, you may use “NA” for the Statistical Delta if this standard is
being followed.
Continuous N
Quality Actual (Statistical Delta)
Characteristic Based on N Actual
Pre-heat D1
8 runs NA
(run to run)
Pre-heat D1
8 runs NA
(within run)
Note: Sample sizes for binomial QCs should follow binomials
methods for defining detectable delta.
If sample sizes are varied, it could bias the standard deviation tests.
Consult a statistician.
SE to Target -or-
Run-Run SEB to POR
SEB to POR
Use “SE to POR (Run to Run), SEB to POR (Within Run)” for non-targeted QCs.
– Note: can conclude SB for non-targeted QCs
Use SEB to qualify new tool, cost reduction projects or productivity improvement
projects.
Use SB for all process improvement projects like reducing non-wet, decreasing Bin15
fallout rate, etc.
CE! Experiments
KPP w/ targets:
SE to Target/SEB to POR
For run-run/within run
Binomial comparisons
Follow standard assembly flow with scripted noise factors for lid and preforms as shown on
scripted table:
After dummy tray run on the tool:
– Take dummy tray and measure on HPR tool and transfer data to station controller.
– Take dummy tray and measure data using scale and send to scale station controller.
Answer for Lisa: Proposed value should include both present and proposed values.
Lisa Ans: Description of experiment is CE! Audit. Therefore only CE! Compliance should be checked, there is no quantity of
material and accept criteria is 100% CE!
Ans for lisa: these are continuous QCs and are listed as binomial. Need to move
them up to continuous QC section. Statistical Delta should be filled out as N/A since
it falls within the 8-16 lot BKM.
1. For a lot level quality characteristics that has a reachable target, what is the accept
criteria for each test. (Possible options: SE to POR, SE to Target, SB to Target, SEB
to POR, SB to POR)
1. Run to Run (Mean) Test – SE to Target
2. Run to Run (Standard Deviation) Test – SEB to POR
3. Within Run (Mean) Test – SEB to POR
4. Within Run (Standard Deviation) Test – SEB to POR
2. Matching experiments with lot level parameters should ideally have at least __ lots (8
is the answer)
Experiment ID:
Objective of the experiment/procedure:
Ensure
Ensure IHS
IHS tool
tool in
in VF
VF is
is matched
matched to
to TD
TD reference
reference tool
tool on
on critical
critical and
and key
key process
process parameters
parameters
Experimental Design
One Sample
Concurrent (including matched pair)
Non-Concurrent
MCA
Other
Details
Execution/DOE Details
7. Process Factors
X
X Placement
Placement Offset
Offset (Run
(Run to
to Run)
Run) ?
X
X Placement
Placement Offset
Offset (Within
(Within Run)
Run) ?
Lid
Lid Pull
Pull Force
Force 120
120 Units
Units
Intel Confidential