SEMI E124-1103



SEMI E124-1103

GUIDE FOR DEFINITION AND CALCULATION OF OVERALL FACTORY EFFICIENCY (OFE) AND OTHER ASSOCIATED FACTORY-LEVEL PRODUCTIVITY METRICS

Purpose

This guide describes metrics that show how well a factory is operating compared to how well it could be operating (for the given product mix). These metrics can be used for tracking factory performance (in value-added production) in a way that rewards good operational decisions and that is not easy to adversely manipulate. The metrics can be used in a process of ongoing improvement that can be visible to all levels of a semiconductor manufacturing organization.

The metrics in this guide are intended for evaluating the relative efficiency of factory production after the factory is in production, not for capacity analysis while the factory is being designed or redesigned. However, some of these metrics can be used in factory simulations for choosing equipment sets and scheduling policies.

Scope

To evaluate the overall effectiveness of a factory, there are at least three things in need of measurement: production, utilization of assets, and costs. This guide focuses on evaluating production. Utilization of assets and costs (as well as other economic factors) are outside its scope.

See ¶ R3-1.1 in Related Information 3 for a discussion of metrics in other areas.

This guide describes metrics for an entire integrated production line (so, for example, lots may not leave and then return to the production line). Multiple production lines in the same factory may be evaluated separately if they do not share resources (such as material handling or production equipment).

NOTICE: This standard does not purport to address safety issues, if any, associated with its use. It is the responsibility of the users of this standard to establish appropriate safety and health practices and determine the applicability of regulatory or other limitations prior to use.

Limitations

In the context of this guide, it is important to note that factory-level productivity is impacted greatly by factors beyond the factory itself, including material availability, efficiency of product device designs, and customer demand.

The metrics in this guide are intended for evaluating the overall efficiency of factory production, not for diagnosing problems (or opportunities for improvement) in the factory, although some component metrics can be used that way.

See ¶ R3-1.2 in Related Information 3 for a discussion of the differences between the two kinds of metrics.

This guide provides metrics and calculations for measuring the overall productivity only of manufacturing environments (such as wafer fabs, flat panel factories, and some disk-drive production facilities) in which product substrates move through the factory with no assembly or disassembly processes. These metrics can be applied in a post-wafer back-end chip production facility if the lead frames are considered to be consumable materials (not units of production in their own right). However, in the future this guide may be extended to comprehend other, more complex manufacturing environments (including assembly operations). Also, additional supplemental metrics may be defined, and the metric definitions given here may be expanded to truly comprehend assembly operations. In addition, metrics may be defined that measure fab productivity from the point of view of wafer movement.

Referenced Standards and Documents

SEMI Standards

SEMI E10 — Specification for Definition and Measurement of Equipment Reliability, Availability, and Maintainability (RAM)

SEMI E79 — Specification for Definition and Measurement of Equipment Productivity

NOTICE: Unless otherwise indicated, all documents cited shall be the latest published versions.

Terminology

Abbreviations and Acronyms

CONWIP — Constant Work in Process

COO — Cost of Ownership

GUE — Good Unit Equivalents

OEE — Overall Equipment Efficiency

OFE — Overall Factory Efficiency

THT — Theoretical Production Time

WIP — Work in Process

Definitions

All the metrics defined below should be calculated with respect to the period being measured.

Several of the metrics defined below are more completely defined in Equations 1 through 22 in ¶ 6.3.

actual throughput rate — the finished units out divided by the total time (shows how fast finished wafers flow out of the factory).

availability efficiency (time divided by time) — the fraction of total time that the equipment is in a condition to perform its intended function (SEMI E79).

average cycle time — the (unweighted) average of cycle time over all of the units of production in finished units out.

average work in process (WIP) — the average cycle time multiplied by the actual throughput rate (shows how many eventually finished units of production fill the “pipeline” on average).

This metric is not an average of the WIP over time because such an average would include units that are later scrapped before finishing.

balance efficiency — the critical WIP divided by the process capacity (measures how well the equipment sets are balanced).

best-case cycle time — the larger of the theoretical cycle time and the quotient of the average WIP divided by the bottleneck throughput rate (shows the best cycle time that the factory can do given the WIP loading).

best-case throughput rate — the smaller of the bottleneck throughput rate and the quotient of the average WIP divided by the theoretical cycle time (shows the best throughput rate that the factory can do given the WIP loading).

bottleneck throughput rate — the upper bound on the factory throughput rate imposed by the current bottleneck equipment set. If a process change for a product causes this metric to change, the product before and after the process change should be considered different products for the purposes of performing these computations.

This metric is similar to (but not the same as) the theoretical unit throughput by recipe metric from SEMI E79.

This metric is not an average over the bottleneck throughput rates of each product.

critical WIP — the theoretical cycle time multiplied by the bottleneck throughput rate (gives the WIP level that theoretically allows the factory to have the highest throughput rate with the shortest cycle time).

cycle time — the amount of time a unit of production spends as WIP in the factory.

finished units out — the number of units of production that finish processing and testing during the period being measured.

good unit equivalents (GUE) out — the (possibly non-integer) number of units of production required to contain all of the good product that exits the factory during the period being measured.

line yield — the fraction of units leaving the factory that have finished processing (measures relative material losses such as scrapped units).

normalized production efficiency — the production efficiency to the power of the normalizing exponent (measures the normalized efficiency of the process with respect to factory dynamics).

normalizing exponent — power that normalizes the production efficiency so that a value of ½ for normalized production efficiency indicates that the factory is performing at the level of the threshold case (which differentiates a well-run factory from one badly operated).

operational efficiency (time divided by time) — the fraction of equipment uptime that the equipment is processing actual units (SEMI E79).

overall equipment efficiency (OEE) (time divided by time) — a metric of equipment performance, expressing the theoretical production time for the effective unit output divided by the total time (CSM 21).[1]

overall factory efficiency (OFE) — the volume efficiency multiplied by the yield efficiency (shows how well a factory is operating compared to how well it could be operating for the given product mix).

process capacity — the maximum number of units of production that can be processed simultaneously throughout the factory (including units being transported by material handling vehicles).

production efficiency — the throughput-rate and cycle-time efficiency multiplied by the WIP efficiency (measures the efficiency of production with respect to factory dynamics).

quality efficiency (time divided by time) — the theoretical production time for effective units divided by the theoretical production time for actual units (SEMI E79).

scrapped units out — the number of units of production (including broken units, external rework, etc.) that exit the factory without finishing production during the period being measured.

set of bottleneck equipment (Fe*) — the collection of production equipment of the same type that has the highest average operational efficiency in the factory during the period being measured. Elements of this set are indicated by “f”, and the equipment type is indicated by “e*”.

This set of bottleneck equipment might not be the equipment set (often the expensive lithography exposure equipment) that was planned to be the bottleneck in the factory, but rather the equipment set with the highest average operational efficiency (the fraction of time in use when available) during the period being measured. If another equipment set experiences significantly lower availability than expected, it might become the bottleneck. Thus, the sets of bottleneck equipment may be different between two adjacent time periods, and the set of bottleneck equipment for the period combining the two adjacent periods may be different from the other two sets.

set of equipment of type e (Fe) — the collection of production equipment of type e(E in the factory. Elements of this set are indicated by “f”.

set of equipment types (E) — the collection of the different types of production equipment in the factory, including metrology equipment and material handling vehicles and conveyors. Elements of this set (which are the different types of equipment) are indicated by “e”.

If units are transported manually between process steps, then the human transporters (and any carts or mechanized vehicles that they operate to perform the movement) should be considered a type of equipment for the purpose of computing the metrics in this guide. This categorization is not intended to dehumanize people, but to ensure that the manual transport time is included in such metrics as theoretical cycle time.

set of process steps of product type p on equipment type e (Spe) — the collection of the different process steps (including metrology inspection and material handling transport) planned for a unit of production of product type p on equipment of type e in the factory. Elements of this set are indicated by “s”.

set of product types (P) — the collection of the different types of products manufactured in the factory. Elements of this set are indicated by “p”.

test yield — the fraction of units leaving the factory that have finished processing and have passed final testing (measures relative losses due to parametric or functional failure).

theoretical cycle time — the minimum time required to process a unit of production through the factory (including material handling transport time) if the unit never has to wait for equipment or a vehicle to become available and if sequence-dependent set-ups never have to be performed. This metric is also known as the raw process time. If a process change for a product causes this metric to change, the product before and after the process change should be considered different products for the purposes of performing these computations. If more than one product (or process flow) is represented in the output, an average is taken over each of the products’ theoretical cycle time weighted by the fraction of that product found in finished units out.

This metric is similar to (but not the same as) the theoretical production time per unit metric from SEMI E79.

theoretical production time per unit (THT) (time per unit) — the minimum rate of time per unit to complete processing, given the following:

• The specified recipe

• The equipment design

• Continuous operation

• No efficiency losses (SEMI E79)

theoretical throughput rate — the smaller of the bottleneck throughput rate and the quotient of the WIP capacity divided by the theoretical cycle time (gives an unreachable upper bound on the factory throughput rate).

theoretical unit throughput by recipe (THTP) (units per time) — for a given production recipe, the number of units per period of time that theoretically could be processed by the equipment. For each recipe, theoretical unit throughput is equal to the reciprocal of theoretical production time per unit (SEMI E79).

throughput rate — the number of units of production that pass through a process per period of time.

throughput-rate and cycle-time efficiency — the best-case cycle time divided by the average cycle time (shows the relative performance of the factory with respect to throughput rate and cycle time).

total time — all time (at the rate of 24 hours per day, seven days per week) during the period being measured. In order to have a valid representation of total time, all six basic equipment states must be accounted for and tracked accurately (SEMI E10).

For factory-level productivity metrics, total time should be larger than the average cycle time (and is recommended to be twice as large as the average cycle time and larger than the cycle time of any individual unit in finished units out).

unit (of production) — the basic entity in the factory (such as a wafer in a fab, a glass pane in a flat panel factory, or a die in a post-wafer back-end chip production facility) that acts as a product substrate (and moves through the factory with no assembly or disassembly processes). Only product units are included (as opposed to test wafers or other non-product devices).

This definition is more restrictive than that given in SEMI E10 in order to be sufficiently specific.

Production lot sizes can (and typically do) include multiple units of production, and units can (and typically do) contain multiple product devices (usually of the same type but possibly of different types). The user may chose to have the production lot be the unit of production, but this choice is not recommended because

• the choice of lot size (and its inherent waiting time while its individual units are serially processed) would no longer be a relevant factor in evaluating how well the factory is running;

• lots can vary in size (even in the same factory);

• the meaning of unit would be inconsistent with other SEMI standards;

• scrapped units out would not be properly accounted for.

uptime (equipment uptime) — the hours when the equipment is in a condition to perform its intended function. It includes productive, standby, and engineering time, and does not include any portion of non-scheduled time (SEMI E10).

volume efficiency — the normalized production efficiency times the balance efficiency (measures the total efficiency of the process with respect to factory dynamics).

WIP capacity — the maximum number of units of production the factory can contain (including on shelves, in stockers, on material handling transport vehicles, on equipment load ports, in internal carrier buffers, and in process chambers, but not including space required for non-product units such as test wafers, dummy wafers, and monitor wafers).

This value is not a practical WIP level because it represents total gridlock of the factory.

WIP efficiency — the quotient of the smaller of the critical WIP and the average WIP divided by the larger of the two (measures the efficiency of WIP levels with respect to factory dynamics).

WIP turnover — the finished units out divided by the average WIP (shows how often the inventory was replaced during the period being measured).

work in process (WIP) — the number of units of production that have been released into the factory but have not yet been scrapped, have not been sent out for external rework, and have not finished processing through all of the production steps.

yield efficiency — the line yield times the test yield (shows overall material efficiency).

This metric is similar to (but not the same as) the quality efficiency metric from SEMI E79.

Symbols

E — set of equipment types

e* — bottleneck equipment type

f — element of an equipment set

Fe — set of equipment of type e

Fe* — set of bottleneck equipment

p — element of a set of product types

P — set of product types

s — element of a set of process steps

Spe — set of process steps of product type p on equipment type e

Calculated Metrics

Figure 1 shows how the metrics defined in § 5 and 6feed into each other. Arrows go from subordinate metrics to the metric in which the subordinate metrics are cited as a part of the primary definition. Shown in the top row (in red rounded rectangles) are the basic building-block metrics for which no equations are needed in this guide; the remaining metrics have equations in this guide that are indicated by the equation numbers in the figure. At the very bottom is the OFE metric into which almost every other metric feeds, although many of the subordinate metrics are useful by themselves(especially if data availability or reliability is a problem). Along the left side (in green rounded rectangles) are the quality metrics that show the efficiency of the process with respect to use of materials. The remainder of the metrics (in blue rectangles) are production metrics that show the efficiency of the process with respect to factory dynamics (without the effects of yield and scrap losses).

Related Information 1 gives an exposition of the underlying science behind these production metrics.

Unlike overall equipment efficiency (OEE), OFE and its factors are not dimensioned in time divided by time, because not all equipment in the factories is present or operating for the same amount of time. Similar to OEE, this metric

• is dependent on product mix, process flow, operations, and time period. These dependencies should be considered when comparing different factories or even comparing different time periods in the same factory when the product mix or process has changed (although such comparisons are still valid), especially when the factory variability is due to external factors (such as demand or excess capacity in non-bottleneck equipment);

• does not comprehend down-stream demand or the varying importance of different products (which might be addressed by a separate metric);

• varies between zero (total chaos or gridlock) and one (unobtainable perfection);

• is a product of dimensionless efficiencies.

[pic]

Definition Tree for Factory-Level Productivity Metrics

Equations 1 through 22 define the remaining metrics not completely defined in § 5.

[pic] (1)

[pic] (2)

[pic] (3)

[pic] (4)

[pic] (5)

[pic] (6)

[pic] (7)

[pic] (8)

[pic] (9)

For x > 0, log2(x) = log10(x) / log10(2) = ln(x) / ln(2).

See ¶ R1-1.7 and ¶ R1-1.8 in Related Information 1 for a derivation of the mathematical expression for the normalizing exponent and for a discussion of the meaning of the threshold case (which differentiates a well-run factory from one badly operated and at which the normalizing exponent causes the normalized production efficiency to have a value of ½).

[pic] (10)

[pic] (11)

The symbols e, e*, E, f, F, p, P, s, and S are defined in ¶ 5.2.23 through ¶ 5.2.27.

[pic] (12)

[pic] (13)

See ¶ R1-1.6 in Related Information 1 for a discussion of this metric.

[pic] (14)

[pic] (15)

[pic] (16)

[pic] (17)

One of the factors in this metric is the average number of available tools in the current bottleneck equipment set, not the total number of tools nominally in the set.

[pic] (18)

See Related Information 1 for why Equations 18 and 20 theoretically represent the best possible cases.

[pic] (19)

[pic] (20)

[pic] (21)

[pic] (22)

Related Documents

SEMI Standards

SEMI E35 — Guide to Calculate Cost of Ownership (COO) Metrics for Semiconductor Manufacturing Equipment

SEMI E58 — Automated Reliability, Availability, and Maintainability Standard (ARAMS): Concepts, Behavior, and Services

SEMI E116 — Specification for Equipment Performance Tracking

NOTICE: Unless otherwise indicated, all documents cited shall be the latest published versions.

MANUFACTURING SCIENCE BACKGROUND

NOTICE: This related information is not an official part of SEMI E124 and was derived from work by the task force. This related information was approved for publication by full letter ballot procedures on July 27, 2003.

To understand the production metrics given in §6 of the main body of this guide, we need to understand the underlying science behind factory dynamics. The most important concept is Little’s Law given in the following equation, which is the same as Equation 15 in §6:

[pic] (R1-1)

In Factory Physics[2], this identity is called the “F = ma” of manufacturing science. Little’s Law relates the three most significant fundamental quantities of production systems. Unfortunately, it says that all three metrics cannot be optimized simultaneously. Little’s Law is shown graphically in Figure R1-1. In all of the figures in this Related Information, the colors denote different values of the normalized production efficiency metric with green representing values close to one (at the bottleneck throughput rate and theoretical cycle time), yellow representing values close to ½ (the threshold case as discussed in ¶ R1-1.7), and red representing values close to zero (when the throughput rate goes to zero or the cycle time gets large). On the floor of Figure R1-1 are the linear contours of the cycle time level sets. Note that the boundaries of the operating region are determined by the theoretical cycle time (Tmin), the bottleneck throughput rate (Rmax), and the WIP capacity (Wmax).

[pic]

Surface Plot of Little’s Law

[pic]

Plot of actual throughput rate vs. average WIP

Looking at two of these fundamental quantities at a time can clarify the factory dynamics. For example, Figure R1-2 above shows actual throughput rate as a function of average WIP levels. Here the diagonal solid black lines represent different constant cycle times, but no known strategy will keep the factory operating exactly on one of these lines.

Now suppose the factory is managed with a push strategy where a constant throughput rate is enforced so that WIP levels are allowed to reach their equilibrium state. As shown below in Figure R1-3, this strategy amounts to choosing to operate the factory on one of the diagonal solid black lines (each of which represent different constant throughput rates). We try to drive the factory along the diagonal line toward the bottom left (for lower average cycle time and average WIP levels) by using better operating principles, but we are resisted by the inherent variability of the factory.

[pic]

Plot of average cycle time vs. average WIP

Now suppose the factory is managed with a pull strategy where a constant WIP level is enforced so that throughput rates are allowed to reach their equilibrium state. In general, this strategy is a better one because studies have shown that a constant WIP level will result in a higher average throughput rate than the constant throughput rate that results in the same average WIP level. As shown below in Figure R1-4, this constant WIP strategy (known as CONWIP) amounts to choosing to operate the factory on one of the solid black curves (each of which represents a different constant WIP level). We try to drive the factory along the curve toward the bottom right (for lower average cycle time and a higher actual throughput rate) by using better operating principles, but we are resisted by the inherent variability of the factory.

[pic]

Plot of average cycle time vs. actual throughput rate

The following derivation shows why the cycle time efficiency (expressed as best-case cycle time divided by average cycle time) and the throughput rate efficiency (expressed as actual throughput-rate divided by best-case throughput-rate) can both be measured by the same metric (throughput-rate and cycle-time efficiency). The derivation also gives alternative definitions for throughput-rate and cycle-time efficiency for use when cycle time information is not available (such as in resource-based simulations).

| | |

|[pic] |(R1-2) |

| | |

The production efficiency is normalized by the power of the normalizing exponent so that a value of ½ for the normalized production efficiency indicates that the factory is performing at the level of the threshold case (which differentiates a well-run factory from one badly operated). This threshold case is also known as the practical worst case because it represents what the best operating procedures can do in a maximally random factory (see Factory Physics for more on this case). In the threshold case,

| | |

|[pic] |(R1-3) |

| | |

which results in the following production efficiency:

| | |

|[pic] |(R1-4) |

Thus, if we set normalized production efficiency to have a value of ½ at this average cycle time, we get the following equation:

| | |

|[pic] |(R1-5) |

| | |

Taking logarithms of both sides,

| | |

|[pic] |(R1-6) |

| | |

we get the following equation:

|[pic] |(R1-7) |

| | |

This last expression is the same as in Equation 9 in §6 of the main body of this guide.

EXAMPLE APPLICATION

NOTICE: This related information is not an official part of SEMI E124 and was derived from an example developed by the task force using this guide. This related information was approved for publication by full letter ballot on April 11, 2003.

As an example of how the metrics in this guide are applied, the diagram in Figure R2-1 shows the process flow for a grossly simplified model of a wafer fab (developed by Karl Kempf at Intel) that has only five machines in three tool sets, one process flow with six steps, and a single material handling vehicle (that transports 25 wafers at a time).

[pic]

Figure R2-1

MiniFab Process Flow

The data for this model are shown in the first few columns of Table R2-1 and Table R2-2. A simulation gave the following additional run data:

total time = 9 years = 4,733,640 minutes finished units out = 938,571 wafers

average cycle time = 1.8 days = 2,592 minutes scrapped units out = 46,929 wafers

total units out = 39,420 lots = 985,500 wafers good unit equivalents out = 891,642.1 wafers

In the next-to-last column of Table R2-1, we added together (for each step) all of the times (to load, process batch, unload, and travel to next step) and summed the results to get a theoretical cycle time of 812.4 minutes. In the last column of Table R2-1, we computed the theoretical production time per unit for each step. These values were then used in the last column of 0to compute the throughput rate for each equipment set. The throughput rate for the lithography equipment set (0.2182 wafers/minute) is the bottleneck throughput rate, not because it is the smallest throughput rate (It is.), but because the lithography equipment set has the highest average operational efficiency. The remaining metrics are computed on the following pages.

Table R2-1 Process Data

|Process Step Number |Equipment |

| |Set |

| |Name |

|[pic] |(R2-1) |

| | |

|[pic] |(R2-2) |

| | |

|[pic] |(R2-3) |

| | |

|[pic] |(R2-4) |

| | |

|[pic] |(R2-5) |

| | |

|[pic] |(R2-6) |

| | |

|[pic] |(R2-7) |

NOTE 17: For this average WIP level, the best-case cycle time was not determined by the theoretical cycle time, but by the bottleneck throughput rate. Thus, a simple metric like (average cycle time)/(theoretical cycle time) underestimates how well the factory is doing.

| | |

|[pic] |(R2-8) |

| | |

|[pic] |(R2-9) |

| | |

|[pic] |(R2-10) |

| | |

|[pic] |(R2-11) |

| | |

|[pic] |(R2-12) |

| | |

|[pic] |(R2-13) |

| | |

|[pic] |(R2-14) |

| | |

|[pic] |(R2-15) |

| | |

|[pic] |(R2-16) |

| | |

|[pic] |(R2-17) |

| | |

|[pic] |(R2-18) |

| | |

|[pic] |(R2-19) |

| | |

|[pic] |(R2-20) |

SUPPLEMENTARY INFORMATION

NOTICE: This related information is not an official part of SEMI E124 and was derived from work by the task force. This related information was approved by full letter ballot procedures on April 11, 2003.

As mentioned in ¶ 2.1, there are at least three things in need of measurement in a factory: production, utilization of assets, and costs. This guide focuses on evaluating production; utilization of assets and costs (as well as other economic factors) are outside its scope. For measuring effectiveness of asset use, an average (over all of the equipment in the factory) of OEE weighted by COO can be defined in some other document. However, the use of consumables, utilities, and human resources would still need to be comprehended. For measuring costs, some other new standard might define a new metric (similar to COO) for the cost of factory ownership expressed in such metrics as $/(good wafers), $/(good chips), $/(metal levels), $/(good transistors), $/circuit, or $/bit.

As mentioned in ¶ 3.2, the metrics in this guide are intended for evaluating the overall health of the factory production, not for diagnosing problems (or opportunities for improvement) in the factory, although some component metrics can be used in that way. These metrics should indicate whether the factory is running poorly (much like taking a person’s temperature indicates whether the person is ill) while some of its components and other diagnostic metrics might indicate what the cause of the problem is (much like doing a blood analysis in the lab identifies the disease). The following are examples of metrics not in this guide that can be used for diagnosis:

• Ratio of turns to WIP at key operations or for blocks of operations

• Overall WIP distribution

• Daily starts and output

• Defect density

• Throughput, utilization, and available up-time of bottleneck equipment

NOTICE: SEMI makes no warranties or representations as to the suitability of the standards set forth herein for any particular application. The determination of the suitability of the standard is solely the responsibility of the user. Users are cautioned to refer to manufacturer's instructions, product labels, product data sheets, and other relevant literature, respecting any materials or equipment mentioned herein. These standards are subject to change without notice.

By publication of this standard, Semiconductor Equipment and Materials International (SEMI) takes no position respecting the validity of any patent rights or copyrights asserted in connection with any items mentioned in this standard. Users of this standard are expressly advised that determination of any such patent rights or copyrights, and the risk of infringement of such rights are entirely their own responsibility.

REVISION RECORD

NOTICE: The following information is provided to track revisions to this document. Negative votes may not be cast against this information. Changes can be submitted to SEMI staff via a Publication Improvement Proposal (PIP) form.

|Cycle |Authorization |Section |Description |

| | | | |

-----------------------

[1] CSM 21: 95-03 Closed-Loop Measurement of Equipment Efficiency & Capacity, 1995; Engineering Systems Research Center, University of California, Berkeley. Tel: (510) 642-4994. Fax: (510) 643-0966. Web site: .

[2] Wallace J. Hopp and Mark L. Spearman, Factory Physics: Foundations of Manufacturing Management, 2nd ed. (New York, NY: Irwin McGraw Hill, 2001).

-----------------------

DRAFT

Document Number: 4368

Date: 5/30/2007

LETTER (YELLOW) BALLOT

Informational (Blue) Ballotæõ

* ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download