1 managing multi-chamber tool productivity or seminar presentation teacher: pros. 陳茂生, pros....
Post on 22-Dec-2015
220 views
TRANSCRIPT
1
Managing Multi-chamber Tool Productivity
OR Seminar PresentationTeacher: Pros. 陳茂生 , Pros. 阮約翰Student: 937807 張幼蘭 2005/4/21
Bruce Auches, Gulsher Grewal, Peter Silverman
Intel Corporation Santa Clara, Ca.
This paper appears in: Advanced Semiconductor Manufacturing Conference and Workshop, 1995. ASMC 95 Proceedings. IEEE/SEMI 1995
Publication Date: 13-15 Nov. 1995 Page(s): 240 – 247
2
Introduction
Because the operational economic benefits, Multi-chamber tools became popular nearly in a decade.
Used in Thin-film, Etching,Testing fields.
3
Measurements of Productivity Run Rate: output wafer per hour (wph) Run/PM Cycle
Motivation: help tool user to make decision of repairing or ignoring failure chamber that maximize productivity.
Various run scenarios
PM
PM Time
Run/PM cycle
4
Delimit Problem Boundary (1/3)
Focus on parallel configuration.
5
Delimit Problem Boundary (2/3)
Parallel processing mode can run by other available chambers.
6
Delimit Problem Boundary (3/3)
Scheduled PM is triggered by fixed processing wafer quantity. Quantity base PM: metal deposition,
poly etching… Time base PM: photo exposure
7
Responses of Unexpected Chamber Failure Incident (1/3)
Full Cluster Operation (FCO): take down tool completely to repair failure chamber.
Necessary if: Central component fail Cannot repair while the rest tool run
Full Run
Full Run
Take down to repair
Chamber Failure
PM
PM Time
Run/PM cycle
8
Responses of Unexpected Chamber Failure Incident (2/3)
Partial Cluster Operation (PCO): defer to repair failure chamber and keep good chambers running until next PM.
Necessary if: Cannot repair while the rest tool run
Full Run
Defer repair and keep Partial Run
Chamber Failure
PM
PM Time
Run/PM cycle
9
Responses of Unexpected Chamber Failure Incident (3/3)
Run/Repair Operation (RRO): repair failure chamber while the rest tool runs.
May or may not be feasible depending on safety issue and failure position.
Full Run
Full Run
Repair with Partial Run
Chamber Failure
PM
PM Time
Run/PM cycle
10
Fixed Variables (1/2)
Tool Run Rate Full Cluster Run Rate (FCRR) Partial Cluster Run Rate (PCRR) As FCRR decreases, FCO is favored.
Mean Wafers between PM (MWBPM) Visiting wafer quantity between PM for
each chamber. As MWBPM increases, FCO is favored.
11
Fixed Variables (2/2)
Major PM Duration (tPM) As long as one chamber finished
MWBPM wafers, major PM is triggered.
As tPM decreases, FCO is favored.
Number of Process Modules (n) Count “Parallel Path” As n decreases, FCO is favored.
12
Failure-dependent Variables
Time to Repair (MTTR) Duration of repairing failure chambers As MTTR decreases, FCO is favored.
Wafer Count (%F * MWBPM) Processed wafers quantity before
chamber failed. As %F increases, FCO is favored.
13
Output Evaluation Formulas(1/4)
W = number of wafers processed in a complete “run/PM” cycle
C = total time in a “run/PM” cycle Output = W/C Higher output is favored
14
Output Evaluation Formulas(2/4)
FCO WFC = MWBPM * n CFC = tBFFC + MTTR + tAFFC + tPM
tBFFC: Time before failure
tBFFC = (%F * MWBPM * n) / FCRR tAFFC: Time after failure
tAFFC = ( ( 1 - %F ) * MWBPM * n) / FCRR
15
Output Evaluation Formulas(3/4)
PCO WPC = WBFPC + WAFPC
WBFPC = MWBPM * n * %F WAFPC = MWBPM * ( n–1 ) * ( 1- %F ),
assume one chamber/path fail for example.
CFC = tBFFC + tAFFC + tPM
tBFFC = (%F * MWBPM * n) / FCRR tAFFC = ( ( 1 - %F ) * MWBPM * (n-1) ) /
PCRR
16
Output Evaluation Formulas(4/4)
RRO WRR = WBFRR + WDFRR + WAFRR
WBFRR = MWBPM * n * %F WDFRR = MTTR * PCRR
If MTTR is long enough that other good chambers/paths reach PM, then WDFRR = WAFPC.
WAFRR = [ MWBPM - WBFRR/n - WDFRR/(n-1) ] * n
CFC = tBFRR + MTTR + tAFRR + tPM
tBFRR = (%F * MWBPM * n) / FCRR tAFRR = WAFRR / FCRR
17
Example (1/3)
Values for variables: n = 2 FCRR = 20 wph (wafers per hour) PCRR = 10 wph tPM = 10 hr (hours) MWBPM = 500 wafers per
chamber/path %F = 20% MTTR = 10 hr
18
Example (2/3)
Output calculation: FCO: 1000 wafers / 70 hr = 14.3 wph PCO: 600 wafers / 60 hr = 10.0 wph RRO: 900 wafers / 60 hr = 15.0 wph
RRO is the best decision if it is feasible; otherwise, FCO is suggested to choose.
Deferring repair would cause 30% of FCO output loss and 50% of RRO output loss.
19
Example (3/3)
RRO Total
Time (hr) 0~10 10~20 20~30 30~40 40~50 50~60 60
Tool State Full Run Repair withPartial Run
Full Run Full Run Full Run PM
Run Rate 20 10 20 20 20 0Wafer Count 200 100 200 200 200 0 900
Output = 15.0
FCO Total
Time (hr) 0~10 10~20 20~30 30~40 40~50 50~60 60~70 70
Tool State Full Run Take down toRepair
Full Run Full Run Full Run Full Run PM
Run Rate 20 0 20 20 20 20 0Wafer Count 200 100 200 200 200 200 0 1000
Output = 14.3
PCO Total
Time (hr) 0~10 10~20 20~30 30~40 40~50 50~60 60Tool State Full Run Partial Run Partial Run Partial Run Partial Run PMRun Rate 20 10 10 10 10 0
Wafer Count 200 100 100 100 100 0 600Output = 10.0
20
Sensitivity Analysis (1/4)
At most time, the RRO is the best strategy; PCO become the best when the MTTF is longer than the time of processing (1-%F) wafers. In previous example, the “break-even
point” of RRO and PCO is at %F = 80%; FCO and PCO is at 72%.
21
Sensitivity Analysis (2/4)
22
Sensitivity Analysis (3/4)
Longer MTTR or later failure timing (bigger %F) lead to choose PCO; or else, lead to choose FCO. Using the data in previous example, it
can be plot a “break-even curve” of FCO and PCO corresponding to %F and MTTF. Above the curve PCO should be employed; below the curve FCO should be employed.
23
Sensitivity Analysis (4/4)
24
Conclusion Tools should be designed to enable
the RRO where successfully maximize output in most cases.
PCO availability should be minimized in most cases. The root causes of the premature failures should be aggressively sought out and fixed.
If RRO is not feasible, tool user should calculate the “break-even curve” to help make decision more quickly.
25
Further Study
Multiple multi-chamber tool repair decision process
Other site unbalance problems