Electronics Guide

How the Catalog Is Organized

SEMI publishes its standards as a multi-volume set, currently sixteen volumes, each covering one domain of semiconductor and adjacent manufacturing: 3D-IC integration, equipment automation in both its hardware and software aspects, flat-panel displays, facilities, gases, high-brightness LEDs, MEMS and NEMS, materials, microlithography, packaging, photovoltaics, process chemicals, silicon materials and process control, safety guidelines, and traceability.

Each document carries a letter prefix that identifies its volume, and the letter is the fastest way to know what a designation is about before you look it up:

• S — safety guidelines: the environmental, health, and safety documents, including SEMI S2, S8, S10, S6, S14, and S22.
• E — equipment automation, information, and control: the SECS, GEM, GEM300, and EDA communication standards, plus equipment metrics and carrier handoff.
• F — facilities: utilities, ultrapure water, chemical distribution components, and the voltage-sag immunity standards.
• M — materials: wafer specifications, substrate dimensions, and characterization. Test methods in the silicon materials and process control volume carry an MF designation.
• P — micropatterning: photomasks, reticles, pellicles, and the data formats used in lithography.
• D — flat-panel display: substrate sizes, display test methods, and display-specific equipment documents.
• C — process chemicals: purity grades and specifications for the liquid chemicals used in wafer processing.
• T — traceability: marking and identification of wafers, carriers, and other production units.

Further volumes cover photovoltaics, MEMS, high-brightness LEDs, and three-dimensional packaging, each with its own prefix. The letter does not predict the reader, though: a facilities engineer meets SEMI F47 through the equipment procurement specification, and an automation programmer meets the S-series through the interlock behavior a tool must exhibit.

Reading a Designation

A full SEMI designation is the prefix, the document number, a hyphen, and a four-digit suffix giving the month and year of that revision. SEMI F42-0600 is the June 2000 revision of the voltage-sag test method, SEMI F47-0706 the July 2006 revision of the sag immunity specification, SEMI S6-0707 the July 2007 revision of the exhaust ventilation guideline. Letters may follow the date to mark a later editorial or partial revision, as in SEMI S2-0712a.

Cite the full designation, including the suffix, in any purchase specification or evaluation report. A tool evaluated to one revision of SEMI S2 has not been evaluated to a later one, and that difference surfaces late in a factory acceptance test, when it is expensive to fix.

Document Types

Not every SEMI document imposes requirements in the same way. The catalog distinguishes specifications, which state what a product or interface must provide; test methods, which state how to measure something so that two laboratories obtain the same answer; guides and practices, which recommend rather than require; terminology documents; and safety guidelines, the S-series form.

The safety guideline form shapes how S-series conformance is judged. It pairs a performance goal with stated criteria that constitute one acceptable way to reach it. A design that does not follow the criteria may still satisfy the goal by another route, and a good evaluation records which applies. That is why S-series conformance is reported as findings rather than as a certificate.

How a SEMI Standard Is Written

The work happens in task forces: small groups of volunteers—equipment suppliers, device makers, materials suppliers, laboratories, sometimes utilities or insurers—each owning a document or a coherent group of documents. Task forces report to technical committees organized by subject: environmental health and safety, information and control, physical interfaces and carriers, facilities, materials, micropatterning, and the rest. SEMI operates these committees through regional chapters in North America, Europe, Japan, Korea, Taiwan, China, Southeast Asia, and India, which is how a standard written largely by one region's suppliers still reflects practice elsewhere.

Balloting and the Treatment of Negatives

A new or revised document reaches the industry through a ballot. The draft is circulated for a defined voting period, and voters may accept, reject, or abstain. A rejection must carry a reason, and this is the substantive part of the process: the task force must review every negative vote and either accept the change it requests or find the objection non-persuasive on the record. An objection cannot simply be outvoted and forgotten. Comments short of a formal negative are also collected and answered.

Ballot results are then reviewed at the technical committee level and audited by SEMI's standards oversight body before publication. The audit checks procedure rather than engineering judgment: whether the ballot was properly announced, whether negatives were addressed, whether the published changes match the changes voted on. Technical judgment stays with the volunteers; procedural integrity is centrally enforced. That separation is what makes the output credible to purchasers who did not participate.

Review, Reapproval, and Inactivation

Published documents are reviewed periodically. A review ends in one of three ways: reapproval unchanged, revision, or inactivation because the technology has gone away or another document has absorbed it. SEMI announces reapprovals publicly, as it did for SEMI F47, which tells purchasers that a standard remains current even though its date suffix has not moved. Inactivated standards stay available for reference, since equipment installed years ago was built to them, but they should not be cited in new procurement.

Voluntary Consensus, Contractual Force

SEMI standards are voluntary. No government mandates conformance to SEMI S2 the way regulators mandate electrical safety law, and SEMI operates no certification scheme granting market access. What gives these documents force is commercial: the major device manufacturers write them into purchase specifications.

The mechanism is worth understanding precisely. A fab's equipment purchase specification states that the tool shall conform to SEMI S2 and SEMI S8 at a named revision, that a third-party evaluation report shall be delivered before shipment, and that outstanding non-conformances shall be closed before final acceptance and payment. At that moment the voluntary guideline becomes a contract term. The same happens with SEMI F47: a fab that has counted its annual voltage sags writes ride-through into the specification, and the supplier passes the requirement down to its power supply vendors. F47 anticipates this, suggesting that users apply it when procuring equipment and that equipment makers impose it on component and module suppliers in turn.

The practical consequence is that a supplier negotiates against a customer rather than a regulator. Deviations are possible, but they must be argued, documented, and accepted in writing. Engineers new to the industry sometimes read "voluntary" as "optional" and discover otherwise when a shipment is held at the loading dock.

The S-Series: Environmental, Health, and Safety

The S-series is the part of the catalog with the most direct consequences for people. Semiconductor process equipment concentrates hazards rarely found together: pyrophoric and toxic gases, strong acids and bases at elevated temperature, high-voltage radio-frequency generators, vacuum systems, robotics moving inside enclosures, lasers, ionizing radiation from implanters, and cryogens. A tool may present a dozen at once, in a room where gowning complicates evacuation.

SEMI S2, the Umbrella Document

SEMI S2, the Environmental, Health, and Safety Guideline for Semiconductor Manufacturing Equipment, was first published in 1993 and is the document everything else in the series orbits. SEMI describes it as a minimum set of performance-based environmental, health, and safety considerations for semiconductor manufacturing equipment. It is organized into roughly nineteen major subject areas: documentation, hazard warning labels, safety interlocks, emergency shutdown, electrical design, fire protection, hazardous energy isolation, mechanical design, seismic protection, automated material handlers and robotics, ergonomics, exhaust ventilation, chemicals, ionizing and non-ionizing radiation, lasers, sound pressure, environmental considerations, and the evaluation and reporting requirements around them.

SEMI S2 does not specify all of this itself. It is an umbrella that invokes the more specialized S-series documents where they exist, so an S2 evaluation is in practice an evaluation against S2 plus a family of referenced guidelines.

SEMI S10: Risk Assessment and Risk Evaluation

SEMI S10, the Safety Guideline for Risk Assessment and Risk Evaluation Process, supplies the method underlying the rest. It defines how to identify hazards associated with a piece of equipment, how to characterize the severity of harm and the likelihood of occurrence, and how to combine those into a risk ranking that determines whether a control is required. Its output is a risk assessment document that accompanies the tool and that the fab's own safety organization can audit.

The discipline S10 enforces is the one found in ISO 12100 for machinery and ISO 14971 for medical devices: hazards must be enumerated before they are controlled, controls tied to specific hazards, and residual risk stated rather than assumed away. What S10 adds is a vocabulary and ranking scheme the semiconductor industry has agreed on, so a risk assessment from one supplier is legible to a customer's engineer who has read hundreds of others.

SEMI S8: Ergonomics

SEMI S8, the Safety Guideline for Ergonomics Engineering of Semiconductor Manufacturing Equipment, was first published in 1995 and is frequently specified alongside S2 as a matched pair. Its subject is the physical interaction between technician and tool: reach distances to components needing routine service, forces to open panels or connect fittings, clearances for a gowned worker wearing gloves, display and control heights, lifting and carrying, and the postures maintenance demands in confined spaces.

Ergonomics reads as a soft subject until the cost is counted. Semiconductor maintenance is repetitive and performed thousands of times over a tool's life, and the awkward reaches and heavy lifts S8 targets are the ones that produce musculoskeletal injury. Moving a filter from the back of a chamber to a front-accessible position costs nothing at the drawing stage and a great deal after installation, which is why fabs put S8 in the purchase specification rather than treating it as advice.

SEMI S6: Exhaust Ventilation

SEMI S6, the Environmental, Health, and Safety Guideline for Exhaust Ventilation of Semiconductor Manufacturing Equipment, covers the local exhaust that removes hazardous gases and vapors from a tool's enclosures and delivers them to the facility abatement system. Its concerns are how much flow each enclosure requires, how flow is measured and monitored, what happens when it is lost, how exhaust connections are labeled and separated by chemical compatibility, and how the equipment states its requirements to the facility that must supply them.

The interface between tool and facility is the crux. A tool designed to a flow assumption the fab cannot meet, or connected to a shared duct carrying an incompatible chemistry, is unsafe however carefully its internal design was reviewed. S6 exists to make the assumption explicit and to force the loss-of-exhaust case into the interlock design, so a tool answers a ventilation failure by shutting down its hazardous processes rather than continuing to run.

SEMI S14: Fire Risk Assessment and Mitigation

SEMI S14, the Safety Guideline for Fire Risk Assessment and Mitigation for Semiconductor Manufacturing Equipment, addresses a hazard fabs treat with particular seriousness, because a cleanroom fire threatens not only the tool that burns but the contamination of an entire production module and months of output. S14 gives the method for assessing a tool's fire risk: identifying ignition sources, characterizing combustible materials in the construction and the process, evaluating the enclosure's ability to contain a fire, and determining what detection and suppression are needed.

The materials question drives many design decisions. Wet benches and chemical enclosures are frequently built from polymers chosen for chemical resistance, and those polymers are not always favorable in a fire. S14 assessments therefore reach into the bill of materials in a way other safety documents do not, and they interact with the insurance requirements large fabs impose independently of any SEMI standard.

SEMI S22: Electrical Design

SEMI S22, the Safety Guideline for the Electrical Design of Semiconductor Manufacturing Equipment, is the electrical counterpart to S2's general provisions. It covers the supply connection and disconnecting means, overcurrent protection, conductor sizing and identification, grounding and bonding, enclosure and access protection against live parts, control circuit design, energy isolation for maintenance, and marking. Radio-frequency generators, high-voltage supplies for implanters, and the stored energy in capacitor banks all receive attention, because those are the hazards that distinguish process equipment from ordinary industrial machinery.

The practical point is that S22 covers ground industrial machinery standards also cover. A fab in the United States may expect NFPA 79; a fab in Europe will expect the harmonized machinery-safety standards, including the electrical requirements of the IEC 60204-1 family. S22 was written for equipment that must satisfy several regimes at once, and a global supplier usually documents against more than one.

The Rest of the Series

The S-series extends well beyond the six documents above, and an S2 evaluation typically touches several. SEMI S1 covers equipment safety labels. SEMI S3 covers process liquid heating systems, the heated chemical baths that have historically been a significant fire and burn hazard. SEMI S4 addresses the separation of chemical cylinders inside dispensing cabinets and SEMI S5 the sizing and identification of flow-limiting devices on gas cylinder valves, both limiting how much hazardous material escapes from a single failure. SEMI S7 concerns the qualification of companies and personnel who perform equipment safety evaluations. SEMI S12 covers decontamination before equipment is moved or serviced, and SEMI S13 the safety documentation the supplier must hand to the user.

Later documents cover further ground: SEMI S16 on design for reduced environmental impact at end of life, S17 on unmanned transport vehicle systems, S18 on flammable silicon compounds such as silane, S19 on training for service personnel, S21 on worker protection, S23 on conservation of energy, utilities, and materials, S24 on multi-employer work areas, S25 on hydrogen peroxide storage and handling, S26 on flat-panel display manufacturing systems, S27 on the contents of evaluation reports, S28 on robots and load ports, and S29 on fluorinated greenhouse gas emissions. S13 and S27 are worth singling out: between them they fix what the customer actually receives and what an evaluation report must contain, which is what makes reports from different evaluators comparable.

The Third-Party S2 Evaluation

A SEMI S2 evaluation is performed by an independent laboratory or engineering firm engaged by the equipment supplier. Intertek, TÜV organizations, UL, and specialist safety consultancies all offer the service. The evaluator reviews the design documentation, inspects and tests the equipment against the applicable criteria in S2 and the documents it references, and issues a report.

What the Report Says

The report does not read as a pass or a fail. Each applicable requirement receives one of a small set of responses; Intertek describes four: the equipment conforms to the stated criteria, it conforms to the performance goal by another means, it does not conform, or the requirement is not applicable. The first two are both passing outcomes, and the distinction matters, because "conforms to the performance goal" records that the supplier met the objective by a route the guideline did not anticipate, and the reasoning is part of the report.

Non-conformances are listed with enough detail for the supplier to act on them. An evaluation is iterative in practice: an early assessment produces findings, the supplier redesigns, and the final report is issued against the shipping configuration. Suppliers who commission the evaluation after the mechanical design is frozen pay for that timing in rework.

How the Report Is Used at Acceptance

The report goes to the customer for review and acceptance. A fab's environmental, health, and safety organization reads it, disputes findings it considers understated, and asks for evidence on the non-conformances that matter. Acceptance is normally a precondition for shipment, and closure of the open findings a precondition for final acceptance after installation. The value of the third-party form is that the fab audits an assessment already performed to a common method instead of repeating it on every tool.

Two limits are worth stating plainly. An S2 report is an evaluation against a guideline, not a certification and not a regulatory approval: it does not substitute for CE marking in Europe, for the electrical listing a local authority may require, or for the fab's own compliance with occupational safety law. And it is a snapshot of one configuration. Field modifications, retrofits, and process changes can invalidate findings, which is why fabs run change control over installed equipment rather than relying on a report written before installation.

Equipment Automation: SECS-I, SECS-II, and HSMS

The E-series automation standards are why a modern fab can run hundreds of tools from dozens of suppliers under one factory control system. They define a common language in which a host computer asks a tool what it is doing and tells it what to do next.

SECS-I (SEMI E4)

SECS-I, the SEMI Equipment Communications Standard 1 Message Transfer, dates from 1978 and defines the transport layer: the hardware, physical, and data-link details for carrying messages between equipment and a host, historically over an RS-232 serial link. It handles framing into blocks, the handshake establishing which end may transmit, checksums, and retries. It is slow by any contemporary measure and survives mainly in legacy equipment and back-end operations where a serial link is adequate. New designs use HSMS.

SECS-II (SEMI E5)

SECS-II, the SEMI Equipment Communications Standard 2 Message Content, published in 1982, is the layer that actually matters, and it has proved remarkably durable. It defines the content of messages independently of how they are carried. Messages are organized into streams, identified by an integer from zero to two hundred fifty-five, and within each stream into functions, similarly numbered. Streams group messages by purpose: equipment status, equipment control, material status, alarms, data collection, and so on.

The numbering convention encodes the conversation. An odd-numbered function is a primary message, and the corresponding even-numbered function is its reply. The canonical example is S1F1, "Are You There," to which the equipment answers S1F2, "On Line Data." A bit in the message header, conventionally called the W-bit, indicates whether the sender expects a reply at all, so a primary message may be sent without one when no answer is needed.

SECS-II also defines the data types a message body may carry: ASCII and JIS-8 text, binary, boolean, four-byte and eight-byte floating point, signed and unsigned integers of one, two, four, or eight bytes, and the list, a container that may hold other items including further lists. Every item is self-describing, so a message body is a nested typed tree rather than a fixed record layout, which is what has let the same standard carry data structures nobody imagined in 1982.

HSMS (SEMI E37)

HSMS, the High-Speed SECS Message Services Generic Services standard, arrived in 1994 and replaces SECS-I's serial link with TCP/IP. It defines how a SECS-II message is framed inside a TCP stream, how a session is established and torn down, how link-test messages confirm that a quiet connection is still alive, and how timeouts are handled. Above HSMS the SECS-II message content is unchanged, which is precisely why the migration was tractable: a fab could move a tool from serial to Ethernet without rewriting the application logic on either end.

HSMS is defined in variants that differ in how sessions are managed, the single-session form being the one most commonly implemented for equipment-to-host connections. When engineers say a tool "has a SECS/GEM interface," they mean SECS-II message content over HSMS transport, conforming to the GEM behavior model described next.

GEM and the 300 mm Standards

SEMI E30: The Generic Equipment Model

SECS-II defines a vocabulary but not a grammar. Two tools could both speak SECS-II and still be unusable together, because nothing compelled them to use the same messages for the same purposes. SEMI E30, the Generic Model for Communications and Control of Manufacturing Equipment, published in 1992 and universally called GEM, supplied the missing agreement: the messages, state machines, and operational scenarios that let factory software control and monitor equipment uniformly.

The GEM behaviors an integrator relies on include a communication state model governing how the link is established and re-established; a control state model determining whether the equipment is offline, under local control, or under host control; event reporting, in which the host attaches chosen variables to chosen equipment events and then receives reports automatically rather than polling; alarm management, with the host enabling or disabling individual alarms and receiving set and clear notifications; status and equipment-constant variables the host can read and, where permitted, write; remote commands to start and stop processing; process program, or recipe, management; spooling, which queues messages generated while the host connection is down so no event is lost; a clock service so equipment and host agree on time; and terminal services, which display a message to the operator at the tool.

GEM's value is that a factory host can be written once against these behaviors. Without it, every tool integration is a bespoke project; with it, integration becomes configuration plus a well-bounded set of equipment-specific extensions.

GEM300

The move to 300 mm wafers around 1999 changed the automation problem. At 300 mm, wafers are too heavy for people to handle in open cassettes, and they travel in sealed front-opening pods carried by an automated material handling system. Fabs became genuinely lights-out in their material movement, and the host needed to track material and manage work at a granularity GEM alone did not provide. The response was a group of standards collectively called GEM300, layered on top of GEM.

The principal documents are SEMI E39, Object Services, defining the concepts, behavior, and services common to the object models the others use; SEMI E40, Processing Management, defining the process job that binds a recipe to a specific set of substrates; SEMI E87, Carrier Management, tracking pods, load ports, and carrier states; SEMI E90, Substrate Tracking, following an individual wafer through the equipment; SEMI E94, Control Job Management, grouping process jobs into a unit the host schedules; SEMI E116, Equipment Performance Tracking, reporting busy, blocked, and idle module states for productivity analysis; SEMI E148, Time Synchronization and the TS-Clock object, keeping timestamps comparable across a fleet; and SEMI E157, Module Process Tracking, reporting the start and end of processing within individual modules.

Alongside these sit the physical interfaces. SEMI E84, the Enhanced Carrier Handoff Parallel I/O Interface, defines the hardwired handshake between an automated material handling system and a load port, the signaling that makes unattended pod delivery safe. SEMI E142 covers substrate mapping, and SEMI E95 the human interface for semiconductor manufacturing equipment, so operators moving between tools meet consistent conventions.

Why This Matters for Fab Productivity

A 300 mm fab is a capital investment measured in billions of dollars, and its economics are dominated by utilization. An hour in which a tool is available but has no material is an hour of that capital earning nothing. The automation standards attack that loss from several directions: the host knows every tool's true state without polling, so scheduling runs on current data; material arrives before a tool goes idle; recipes are downloaded rather than selected by hand, removing a common source of misprocessing; and substrate-level tracking identifies the affected wafers exactly when a problem is found.

Measurement is standardized too. SEMI E10 defines equipment reliability, availability, and maintainability metrics together with the state model beneath them, dividing all time into states such as productive, standby, engineering, scheduled downtime, unscheduled downtime, and non-scheduled. Because every supplier reports against the same definitions, a fab can compare tools, hold suppliers to contractual availability commitments, and compute overall equipment efficiency consistently. Without an agreed state model, availability figures from two suppliers are not comparable and every negotiation restarts from definitions.

EDA and Interface A

GEM was designed for control. It answers "what should this tool do next," moving a modest volume of data on a channel that must stay responsive. Advanced process control, fault detection and classification, and equipment health monitoring ask a different question: "what exactly did this tool do, sampled as finely as its sensors allow." A data-collection request must never delay a command, so that traffic does not belong on the control channel.

The Equipment Data Acquisition standards, known collectively as EDA or Interface A, provide a second, read-only interface for the purpose. SEMI describes EDA as defining high-speed data publication from any equipment to any data consumer through web services. The control channel remains GEM; the data channel becomes EDA, and multiple clients may collect from it at once without interfering with one another or with production.

The Standards in the Suite

SEMI E120 defines the Common Equipment Model, a hierarchical description of a tool as modules, subsystems, and components. SEMI E125 defines Equipment Self-Description, by which a tool publishes its own metadata—its modules, the parameters each exposes, the events it can raise—so a client discovers its capabilities at run time rather than being configured with them in advance. SEMI E132 defines authentication and authorization, controlling which clients may connect and what each may see. SEMI E134 defines data collection management, the plans specifying what to sample, how often, and on which trigger. SEMI E128 specifies XML message structures, E145 classifies measurement unit symbols in XML, and E138 covers common components. SEMI E164, the specification for EDA Common Metadata, was added in 2012.

Self-description is the interesting part. A GEM interface requires that the host be told, out of band, which status variable identifier corresponds to which physical quantity. An EDA interface publishes that mapping itself, tied to the equipment model, so a client walks the tool's own description and builds a collection plan against it. That is what makes it feasible to instrument a fleet of heterogeneous tools without a per-tool integration project.

Freezes

Because EDA is a suite of interlocking documents, a tool cannot usefully implement one at a revision that does not match the others. SEMI addressed this with the freeze: a named, mutually consistent set of revisions that suppliers and fabs specify as a unit. One such freeze combines SEMI E120-0310, E125-0710, E128-0310, E132-0310, E134-0710, and E138-0709, and a later generation has been developed under a further freeze. When writing a procurement specification, name the freeze rather than the individual documents; naming them individually invites a combination nobody has tested.

Equipment Cybersecurity

Connecting every tool to the factory network also created a security problem, because fab equipment runs long-lived embedded controllers that cannot be patched on a normal schedule, inside a network holding recipes that are among a manufacturer's most valuable secrets. SEMI answered in 2022 with E187, which sets security requirements for fab equipment covering the operating system, network security, endpoint protection, and security monitoring, and E188, which governs malware-free equipment integration, the process of bringing a tool and its accompanying software, media, and service laptops into the fab without carrying an infection across the boundary. Further documents have followed, including E190 and E191.

SEMI F47 and Voltage-Sag Ride-Through

Of all SEMI standards, F47 is the one most likely to reach an engineer who has never seen a wafer. It is a facilities specification with consequences that run down to individual power supplies.

The Problem

A voltage sag is a decrease in supply voltage below ninety percent of nominal that stops short of a complete interruption, typically lasting three to ten cycles, roughly fifty to one hundred sixty-seven milliseconds. Sags come from faults elsewhere on the distribution network: a car striking a pole, a tree contacting a line, a fault on an adjacent feeder, a large motor starting nearby. They are far more common than outages. Work by the Electric Power Research Institute's Power Electronics Applications Center, cited by Pacific Gas and Electric, concluded that a typical customer could experience about twelve utility voltage sags per year.

A sag of that length is invisible to a person and devastating to a fab. When a tool drops out, the wafers in it are typically scrapped, the tool must be recovered and requalified, and the disturbance propagates through work in progress. The frustrating part is that the equipment usually did not have to fail: a relay dropped out, an undervoltage trip fired, or a lightly specified power supply lost regulation, while the process itself would have continued quite happily.

What F47 Requires

SEMI F47, the Specification for Semiconductor Processing Equipment Voltage Sag Immunity, first published in 2000, defines a required ride-through envelope. Equipment must operate without interruption through sags of the following depths and durations:

• Down to fifty percent of nominal voltage for durations from fifty to two hundred milliseconds, which is three to twelve cycles at sixty hertz.
• Down to seventy percent of nominal voltage for durations from two hundred to five hundred milliseconds, or twelve to thirty cycles at sixty hertz.
• Down to eighty percent of nominal voltage for durations from five hundred to one thousand milliseconds, or thirty to sixty cycles at sixty hertz.

Behavior for sags shorter than fifty milliseconds or longer than one second is not specified as a requirement. The specification adds recommended, but not required, thresholds: tolerance of a sag to zero percent for one cycle, to eighty percent for ten seconds, and continuous operation at ninety percent.

The scope covers semiconductor processing equipment, metrology equipment, and automated test equipment. The companion document SEMI F42, the Test Method for Semiconductor Processing Equipment Voltage Sag Immunity, defines how conformance is demonstrated: test equipment, safety precautions, the processing modes the equipment must be in, the test sequence, phase connections, and reporting. The 2006 revision of F47 references IEC 61000-4-34 for immunity to voltage dips and short interruptions, aligning the semiconductor requirement with the broader electromagnetic compatibility test infrastructure.

How Designers Meet It

The cheapest fixes are specification choices rather than added hardware. Selecting a power supply whose input range places the nominal voltage near the top of that range leaves headroom for a sag. Connecting a single-phase supply phase to phase, where the rating allows and three-phase power is available, gains substantial margin. Loading a supply well below its rating extends hold-up, which depends on the energy stored in the bulk capacitance against the load drawn. A three-phase supply helps because a sag on one or two phases leaves the others supporting the DC bus, and running loads from a shared DC bus backed by capacitance or batteries converts many separate ride-through problems into one.

The remaining failures are usually control devices rather than power converters. Undervoltage relays, phase-loss detectors, and internal reset circuits often trip long before the process would have suffered, and their thresholds and time delays are frequently set conservatively for no considered reason. Contactors can be given more hold-in margin or replaced with devices designed to ride through. Only when these avenues are exhausted does voltage-regulating equipment such as a ferroresonant transformer or an electronic sag compensator make sense, and a mitigation device placed upstream of the actual cause can make matters worse.

Because SEMI F47 is a procurement requirement rather than a law, its effect propagates down the supply chain. Power supply vendors publish F47 test reports, and equipment makers select from those to build a tool whose behavior can be demonstrated. That is a voluntary standard doing what regulation cannot: it made ride-through a specifiable, comparable product attribute.

The M-Series: Wafer and Materials Specifications

The M-series is where the SEMI standards program began, and it remains the least visible and most universally relied upon part of the catalog. SEMI M1, the specification for polished monocrystalline silicon wafers, and its companions define what a wafer physically is: diameter and thickness with their tolerances, edge profile, crystallographic orientation, the orientation feature, surface finish, and geometric parameters such as total thickness variation, bow, and warp that determine whether a wafer will chuck flat and image correctly.

Diameters and Thicknesses

Nominal thickness rises with diameter, because a larger wafer must resist its own weight and the stresses of handling and thermal cycling. The progression runs one hundred millimeters at five hundred twenty-five micrometers, introduced in 1976; one hundred twenty-five at six hundred twenty-five, from 1981; one hundred fifty at six hundred seventy-five, from 1983; two hundred at seven hundred twenty-five, from 1992; and three hundred at seven hundred seventy-five, from 1999. Each value is a standard number rather than a physical necessity, and that is the point: robot end effectors, cassette slot pitches, chuck designs, and process recipes across the industry are built around them.

Orientation Features

Older, smaller wafers carry flats ground into the edge. A primary flat indicates crystallographic orientation, and on earlier generations a secondary flat at a defined angle encoded the conductivity type, so doping could be read visually. Wafers of two hundred millimeters and above use a single small notch instead. The notch consumes far less area than a flat, which matters when the lost region is measured against die yield, and it carries no doping information, which by then was tracked administratively.

Test Methods

Alongside the specifications sit the test methods that make them measurable. Documents in the silicon materials and process control volume, designated MF, define how to determine resistivity, oxygen and carbon content, flatness, particle counts, and the other parameters a wafer specification calls out. Without agreed test methods a specification is unenforceable: supplier and customer measure the same wafer, get different numbers, and cannot resolve the difference. Pairing a specification with a test method is a general principle here, and F47 with F42 is the same pattern in the facilities volume.

SEMI Alongside IEC and ISO

SEMI standards do not exist in isolation. A semiconductor tool shipped internationally is at once a piece of industrial machinery, an item of electrical equipment, and, in the European Union, a product requiring CE marking. Understanding how the regimes relate prevents both duplicated work and dangerous gaps.

Where They Complement Each Other

The clearest division is by specificity. IEC and ISO documents are horizontal: machinery safety, electrical equipment of machines, risk assessment methodology, ergonomics, and electromagnetic compatibility for all industries. SEMI documents are vertical: the hazards and interfaces peculiar to semiconductor manufacturing, which the horizontal standards do not treat in useful detail. No IEC standard tells a designer how to arrange exhaust for a wet bench handling hot sulfuric acid, and no SEMI standard restates the general principles of machinery risk reduction as thoroughly as ISO 12100.

In several places SEMI builds on the horizontal documents rather than duplicating them. The 2006 revision of SEMI F47 points to IEC 61000-4-34 for test methodology, so the sag immunity test uses the same equipment and procedures as general immunity testing. SEMI S22 covers ground that IEC 60204-1 and NFPA 79 also cover, and global suppliers commonly document to more than one. SEMI S10 shares its analytical structure with ISO 12100, so a risk assessment prepared for one purpose is usually adaptable to the other.

Where They Conflict

Conflicts arise less from contradictory requirements than from different framings, and they cause trouble because different actors enforce the two systems.

The first is legal status. In the European Union, machinery must satisfy the essential health and safety requirements of the machinery legislation—historically Directive 2006/42/EC, succeeded by Regulation (EU) 2023/1230—and the manufacturer must affix CE marking and issue a declaration of conformity. A SEMI S2 report is not a substitute; it is engineering evidence a customer values, while the legal obligation is discharged through the harmonized standards route. The reverse holds too: CE marking does not satisfy a fab's S2 requirement, because S2 asks questions about semiconductor-specific hazards that the machinery route never poses.

The second is the criteria-versus-requirements framing. SEMI safety guidelines pair performance goals with stated criteria and allow conformance by an alternative route. Harmonized European standards work differently: departing from the standard is permitted, but the manufacturer then bears the burden of demonstrating equivalence in the technical file. An argument that satisfies an S2 evaluator as "conforms to the performance goal" is not automatically sufficient there, and the two records must be maintained with that difference in mind.

The third is numerical divergence in shared subject matter. Where SEMI and an IEC or ISO document both address a topic—creepage and clearance, guard interlock reliability, ergonomic force limits, sound pressure—the values and measurement conditions may differ, because different committees derived them for different populations of equipment. The answer is neither to pick one nor to average them, but to design to the more demanding value on each parameter and document against both. That costs a little margin and saves a great deal of argument at acceptance.

Working With SEMI Standards

SEMI standards are sold individually and as volume subscriptions through the SEMI store. Because the families are interdependent, buying a single document is often a false economy: an S2 evaluation reaches into a dozen other S-series guidelines, and an EDA implementation requires the whole freeze. Committee participation is open to employees of member companies, and for a supplier whose product is affected by a document under revision it is both the only way to influence the outcome and months of warning about what will reach customer specifications.

Several errors recur often enough to name. Citing a standard without its revision suffix leaves the requirement ambiguous and, in a contract, unenforceable. Treating an S2 report as a certificate or a regulatory approval misrepresents what it is. Commissioning a safety evaluation after the mechanical design is frozen guarantees rework. Specifying EDA documents individually rather than by freeze produces an untested combination. Assuming that a tool advertising a "SECS/GEM interface" implements the full GEM behavior set rather than a subset leads to integration surprises; read the supplier's GEM compliance statement instead. And treating "voluntary" as "optional" ignores how these documents acquire force, which is through the purchase order.

Conclusion

The SEMI standards program shows that an industry can regulate its own interfaces effectively when the commercial incentive is strong enough. Nothing compels a company to follow SEMI S2, SEMI E30, or SEMI F47. The documents govern the industry anyway, because the customers who buy semiconductor equipment write them into purchase specifications and hold shipment until they are met. Voluntary consensus plus contractual adoption comes close to the effect of regulation, with faster revision cycles and technical content written by the people who build the equipment.

For a practicing engineer the map is simple. The S-series determines whether a tool is safe enough to install, enforced through third-party evaluation reports reviewed at acceptance, with SEMI S2 as the umbrella and S6, S8, S10, S14, and S22 supplying depth on ventilation, ergonomics, risk method, fire, and electrical design. The E-series determines whether the tool can be automated and measured, running from SECS-II over HSMS, through GEM and the GEM300 material and job-tracking standards, to the EDA data channel and the E187 and E188 security requirements. The F-series determines what the facility must supply and the equipment must tolerate. The M-series determines what a wafer is.

None of this replaces the horizontal IEC and ISO standards that apply to the tool as machinery and as electrical equipment. The two systems overlap, occasionally disagree on specific values, and are enforced by different parties. Engineers who identify the full applicable set at the start of a program, maintain one hazard register feeding every downstream document, and design to the more demanding value where the regimes diverge spend far less time arguing at acceptance than those who find the conflict when the tool is already on the loading dock.

Related Topics