Skip to content

METRIC CONTRACTS / V132

Semiconductor Metric & Terminology Ontology

Check what a value means before comparing the number.

English master terms and Korean aliases lead to explicit quantity, unit, denominator, sign, grain, condition, and extraction contracts. Similar labels never become equivalent by guesswork.

Canonical terms
21
84 searchable names
Metric contracts
471
384 remain UNKNOWN
Bound surfaces
449
zero unbound
Explicit relations
17
four visible states
01 Identify the contract02 Match every dimension03 Apply only declared conversion04 Refuse what stays unknown

INSPECTION WORKSPACE

Definitions first. Differences second.

Search bilingual terminology, inspect complete and unresolved contracts, then test one explicit relationship without storing or sending your selection.

Loading all 471 contracts locally...

21 / 21 terms
01

Threshold voltage

임계 전압 / 문턱 전압

VTH · Vth

A declared extraction result used to mark MOSFET turn-on under specified terminal bias, geometry, temperature, and method.

It is not one universal physical switch point. Constant-current and transconductance-extrapolation values are different metric contracts.
02

Off-state current

오프 상태 전류 / 오프 전류

Ioff · I_OFF

Drain current measured or computed at a declared off-state bias and temperature.

Leakage current is broader. Total-device current and geometry-normalized current are not interchangeable.
03

Transconductance efficiency

트랜스컨덕턴스 효율 / 상호 컨덕턴스 효율

gm/Id · g_m/I_D

Small-signal transconductance divided by drain current at one declared operating point.

The S/A and inverse-volt forms are dimensionally equivalent only when bias, extraction, and current sign conventions match.
04

Critical dimension

임계 치수 / 주요 선폭

CD

A specified geometric dimension extracted at a declared feature, measurement plane, method, and sampling grain.

CD does not identify top, middle, bottom, drawn, or electrical width by itself.
05

Line-edge roughness

선 가장자리 거칠기 / 라인 에지 거칠기

LER

Statistical variation of one edge position after a declared detrend, bandwidth, sampling, and estimator.

RMS and three-sigma forms require an explicit conversion under the same samples and estimator. LER is not LWR.
06

Line-width roughness

선폭 거칠기 / 라인 폭 거칠기

LWR

Statistical variation of right-edge minus left-edge linewidth after a declared detrend, bandwidth, sampling, and estimator.

Its relationship to LER depends on both edge variances and covariance; two times LER is not universal.
07

Local critical dimension

국부 임계 치수 / 로컬 CD

LCD · local CD

A local linewidth statistic computed over a declared spatial aperture and sampling rule.

Aperture, edge definition, sample count, and histogram denominator are part of the contract.
08

Propagation delay

전파 지연

tpd · t_pd

Elapsed time between declared input and output threshold crossings with direction and interpolation fixed.

Linear interpolation and first-saved-sample extraction are related but not numerically equivalent.
09

Rise time

상승 시간

tr · t_r

Elapsed time for one waveform to cross a lower and upper threshold in the rising direction.

Rise time is not input-to-output propagation delay even when both use picoseconds.
10

Static noise margin

정적 잡음 여유 / 정적 노이즈 마진

SNM

A declared DC stability margin of a bistable cell, commonly extracted from a butterfly construction.

Hold, read, and write contexts are different contracts and must retain cell, bias, and extraction conditions.
11

Read-margin spread

읽기 마진 산포 / 리드 마진 산포

sigma_read

Dispersion of a declared read-margin observable over a named hierarchy and condition set.

Row count is not automatically the independent sample count when lot, wafer, site, cell, condition, or replicate levels are shared.
12

Variance closure

분산 폐합 / 분산 성분 폐합

Var

A ledger relating total variance to component variances and covariance terms under one declared sample set.

Finite subsets and missingness can leave a covariance remainder; it must not be silently forced to zero.
13

Condition coverage

조건 커버리지 / 조건 포함률

coverage

Observed support relative to an explicitly declared condition set and counted unit.

Selected, common, union, and missing-condition denominators are distinct; missing is not numeric zero.
14

Normalized ratio

정규화 비율 / 정규화 분율

r

A dimensionless ratio with an explicitly named numerator, denominator, domain, and grain.

A 0-to-1 teaching proxy is not automatically a probability, yield, confidence, or quality score.
15

Percent

백분율 / 퍼센트

%

One hundred times a declared dimensionless ratio.

The percent symbol does not supply a missing numerator, denominator, population, or uncertainty model.
16

Yield

수율

Y

A good-unit count divided by a declared opportunity or started-unit denominator at a named grain.

Die yield, wafer yield, lot yield, first-pass yield, and modeled yield are not interchangeable.
17

Film thickness

막 두께 / 박막 두께

t

Distance between declared film interfaces measured or computed by a specified method and grain.

Mean, center, edge, local, and fitted thickness need separate context even when units convert exactly.
18

Etch selectivity

식각 선택비 / 에치 선택비

S_A/B

Etch rate of named material A divided by etch rate of named material B under the same declared condition.

Reversing numerator and denominator requires an explicit reciprocal conversion and a nonzero value.
19

Drain-induced barrier lowering

드레인 유도 장벽 저하

DIBL

Change in an explicitly extracted threshold metric per declared drain-bias change.

Signed derivative and positive degradation magnitude use different sign conventions and units.
20

Normalized teaching proxy

정규화 교육용 대리지표

proxy

A packet-specific dimensionless value generated from declared original teaching equations.

Identical labels or 0-to-1 ranges across packets do not establish equivalence, desirability, causality, or calibration.
21

Legacy display metric

기존 표시 지표

legacy

A public metric label and value string whose complete comparison dimensions were not structured in the V131 catalog.

It receives a stable identifier but remains UNKNOWN and cannot enter numeric comparison until its missing fields are reviewed.

PUBLIC EVIDENCE BOUNDARY

Compatibility is not validity.

A structurally comparable pair can still come from a poor measurement, invalid model, or unsupported claim.

  • No inferred equivalenceAliases, symbols, units, and ranges are search aids, not proof.
  • No silent conversionEquations, inverse equations, and assumptions remain visible.
  • No hidden scoreDirection of better, quality, and readiness are never guessed.
  • No production dispositionNo recipe, yield prediction, qualification, or signoff follows.