Electronics Guide

Test Limit Derivation

Go/no-go test limits must be derived from compliance test limits with appropriate margins to account for measurement uncertainty, environmental differences, and production variation. The limit-setting process involves several considerations:

Compliance margin: Production test limits should be tighter than compliance limits to ensure that passing products have adequate margin for compliance testing. Typical practice uses limits 3-6 dB inside compliance limits, depending on measurement uncertainty and production variation.

Measurement system uncertainty: Production test equipment may have greater measurement uncertainty than laboratory equipment due to environmental conditions, fixture effects, and calibration intervals. These uncertainties must be included in the limit calculation.

Correlation offset: Systematic differences between production and compliance test results require correction in the production test limits. Correlation studies quantify these offsets, which may vary with frequency, product configuration, or other factors.

Statistical variation: Production variation causes test results to vary between units. Test limits must account for this variation to achieve the desired balance between false failures (compliant units failing production test) and escapes (non-compliant units passing production test).

Test Parameter Selection

Not all EMC parameters can be practically tested in production, so go/no-go testing typically focuses on parameters that are most indicative of overall EMC performance and most susceptible to production variation:

Radiated emissions: Production radiated emissions tests often focus on frequencies where the product has minimum margin or where production variation has the greatest effect. Broadband scanning may be replaced by spot frequency measurements at critical frequencies.

Conducted emissions: Conducted emissions are more amenable to in-line testing than radiated emissions because they require less elaborate test setups. Line impedance stabilization networks (LISNs) can be integrated into test fixtures for direct connection to the product.

Immunity proxies: Full immunity testing is often impractical in production. Proxy measurements such as power supply rejection ratio, filter attenuation, or shield continuity may indicate immunity performance without requiring the extensive test setups of full immunity testing.

Functional indicators: Some EMC characteristics correlate with functional performance measurements. Clock frequency accuracy, signal rise times, or power supply noise may indicate EMC performance without explicit EMC measurements.

Decision Rules and Actions

Clear decision rules define the actions taken based on test results:

Pass: Products passing all tests continue to subsequent production operations or to finished goods. Pass results are logged for traceability and trend analysis.

Fail: Products failing any test are diverted for diagnosis, rework, or scrap. The failure mode and test data are recorded for analysis. Depending on the failure, products may be retested after rework or may require full compliance testing before release.

Marginal: Some programs define a marginal zone between pass and fail limits. Marginal products may receive additional testing or may be flagged for enhanced scrutiny even if they pass. Marginal results often indicate process drift before it causes failures.

Retest: When results are near limits or when test conditions are questionable, retesting may be appropriate. Retest policies should prevent repeatedly testing the same unit hoping for a passing result while allowing genuine measurement errors to be corrected.

Sampling Plan Design

Sampling plans define how many units to test and how to select them from production. Key elements include:

Sample size: Larger samples provide more confidence in population characteristics but consume more test resources. Statistical methods determine the sample size needed to achieve specified confidence levels and acceptable quality limits.

Sampling frequency: Samples may be drawn at fixed intervals (every Nth unit), at fixed times (once per hour or shift), or triggered by events (after setup, changeover, or maintenance). The appropriate frequency depends on production rate and the expected rate of process changes.

Selection method: Random sampling prevents bias but may miss problems concentrated in specific time periods or positions. Stratified sampling ensures representation from different conditions such as different shifts, machines, or operators.

Acceptance criteria: The sampling plan defines how many failures are acceptable within a sample. Zero-defect plans require all sampled units to pass, while other plans accept a specified number of failures before rejecting the lot.

Lot-Based vs. Continuous Sampling

Sampling strategies fall into two broad categories based on how production is organized:

Lot-based sampling: Production is organized into discrete lots, and samples are drawn from each lot. The lot is accepted or rejected based on sample results. This approach suits production with natural batch boundaries such as different production runs, material lots, or setup conditions.

Continuous sampling: Production flows continuously without natural lot boundaries, and samples are drawn on an ongoing basis. Statistical process control methods track results over time and trigger investigation when results indicate process changes. This approach suits high-volume continuous production.

Hybrid approaches combine elements of both, using lot-based acceptance for material lots while applying continuous monitoring across lots to detect gradual drift.

Switching Rules

Adaptive sampling plans adjust sampling intensity based on quality history:

Normal inspection: The baseline sampling rate applies when quality history is normal and no special conditions exist.

Tightened inspection: When recent results show degradation or when failure rates increase, sampling intensity increases to provide earlier detection of problems and more information for diagnosis.

Reduced inspection: When quality history demonstrates consistently good performance, sampling may be reduced to conserve test resources. The criteria for reduced inspection are typically more stringent than those for normal inspection.

Switching rules define the conditions that trigger transitions between inspection levels, including the number of consecutive lots at each level before switching and the events that force immediate return to normal or tightened inspection.

Acceptance Sampling Standards

Several standards define acceptance sampling procedures for different applications:

ANSI/ASQ Z1.4 (formerly MIL-STD-105): This standard defines attribute sampling plans where each sampled unit is classified as conforming or nonconforming. It provides tables for selecting sample sizes and acceptance numbers based on lot size, inspection level, and acceptable quality limit (AQL).

ANSI/ASQ Z1.9 (formerly MIL-STD-414): This standard defines variables sampling plans where actual measured values are recorded rather than simple pass/fail results. Variables plans require smaller samples to achieve the same discrimination as attribute plans but require normally distributed measurements.

ISO 2859 series: The international equivalents of the ANSI sampling standards provide similar procedures with some differences in terminology and table organization. These are often specified for international programs or when harmonization with international practices is required.

Selecting the appropriate standard and plan involves balancing producer's risk (probability of rejecting good lots) against consumer's risk (probability of accepting bad lots) while considering the practical constraints of sample size and testing cost.

Operating Characteristic Curves

Operating characteristic (OC) curves show the probability of accepting lots as a function of the true lot quality. These curves characterize the discrimination power of sampling plans:

An ideal sampling plan would accept all lots with quality better than the AQL and reject all lots with quality worse than some rejectable quality level (RQL). Real sampling plans have OC curves that transition gradually between acceptance and rejection regions.

Key points on the OC curve include:

• AQL point: The quality level at which lots have a high probability of acceptance (typically 95%)
• RQL point: The quality level at which lots have a low probability of acceptance (typically 10%)
• Indifference quality: The quality level at which acceptance probability equals 50%

Comparing OC curves for different sampling plans helps select plans that provide appropriate discrimination for the application.

Average Outgoing Quality

When rejected lots are subjected to 100% inspection and defective units are removed or repaired, the average outgoing quality (AOQ) differs from the incoming quality. The AOQ curve shows the expected outgoing quality as a function of incoming quality:

The maximum point on the AOQ curve is the Average Outgoing Quality Limit (AOQL), which represents the worst average quality that can reach the customer regardless of incoming quality. This occurs because very bad lots are likely to be rejected and 100% inspected, while very good lots pass sampling with few defects.

Plans can be selected based on AOQL when the customer's concern is the long-term average quality rather than individual lot quality.

Control Chart Methods

Statistical process control charts provide visual and statistical tools for monitoring process stability:

X-bar and R charts: These paired charts track the average (X-bar) and range (R) of small subgroups of measurements. Changes in the process mean appear on the X-bar chart, while changes in process variability appear on the R chart. Together, they provide comprehensive monitoring of process stability.

Individual and moving range charts: When subgrouping is not practical (for example, when testing is destructive or very expensive), individual measurements can be charted with moving ranges calculated from consecutive measurements. These charts are less sensitive than subgroup charts but require only one measurement per time period.

CUSUM and EWMA charts: Cumulative sum (CUSUM) and exponentially weighted moving average (EWMA) charts are more sensitive to small sustained shifts than Shewhart charts (X-bar and R). They accumulate information from multiple samples to detect gradual changes more quickly.

Control limits on these charts are typically set at three standard deviations from the process mean, providing approximately 99.7% probability that points within the limits represent normal process variation.

Out-of-Control Signals

Control charts signal potential problems through several patterns:

Points beyond limits: A single point beyond control limits suggests a special cause affecting that sample. Investigation should identify whether the cause is measurement error, a process upset, or a systematic change.

Run rules: Patterns within the control limits can also signal process changes. Common run rules include:

• Seven or more consecutive points on one side of the center line (shift in mean)
• Seven or more consecutive points trending up or down (trend)
• Two of three consecutive points beyond two standard deviations (shift)
• Fourteen or more consecutive points alternating up and down (stratification or sampling issues)

Non-random patterns: Cyclical patterns, stratification, or clustering suggest systematic causes related to equipment cycles, measurement systems, or environmental factors.

Process Capability Analysis

Process capability indices quantify how well the process meets specifications:

Cp: The process capability index compares the specification width to the process spread (six standard deviations). Cp = (USL - LSL) / 6 sigma. A Cp of 1.0 indicates the process spread just fills the specification range, while higher values indicate more margin.

Cpk: The process capability index Cpk accounts for process centering as well as spread. Cpk = minimum of (USL - mean) / 3 sigma or (mean - LSL) / 3 sigma. A process can have high Cp but low Cpk if it is not centered within the specification range.

Pp and Ppk: Performance indices use overall standard deviation rather than within-subgroup standard deviation. They reflect actual performance including between-subgroup variation that Cp and Cpk exclude.

For EMC applications, capability indices should be calculated relative to internal limits (with compliance margin) rather than just compliance limits, to ensure adequate margin in the final product.

Electrical Interface Design

The electrical interface between fixture and DUT must accurately represent the product's intended operating conditions while providing reliable connections for thousands or millions of test cycles:

Connection technology: Spring-loaded pins (pogo pins), pressure contacts, or edge connectors provide the physical connection to the DUT. Contact technology selection considers the number of connections, required current capacity, high-frequency performance, and expected wear life.

Impedance control: For high-frequency measurements, fixture traces and cables must maintain controlled impedance to prevent reflections and measurement errors. Fixture PCBs often use the same layer stack and trace geometries as the product to maintain impedance consistency.

Grounding structure: The fixture ground system must provide a consistent reference for measurements while avoiding ground loops that could inject interference. Ground plane fixtures with multiple ground connections to the DUT and to the test system minimize ground impedance.

Filtering and isolation: Fixtures may incorporate filtering to prevent test signals from affecting other circuits or to protect sensitive measurements from interference. Relay matrices or electronic switches enable different test configurations without physical reconfiguration.

Mechanical Design Considerations

Mechanical aspects of fixture design affect both test quality and production throughput:

Clamping and alignment: The fixture must position the DUT accurately and repeatably for consistent electrical contact. Alignment features guide insertion, while clamping mechanisms secure the DUT during testing. The clamping force must be sufficient for reliable contact without damaging the product.

Accessibility: The fixture must accommodate product insertion and removal without excessive operator effort or cycle time. Automated fixtures with pneumatic or electric actuation can reduce cycle time and operator fatigue.

Durability: Production fixtures may experience millions of cycles over their lifetime. Contact pins wear, mechanisms fatigue, and alignment drifts. Design for durability includes selecting appropriate materials, providing adjustability, and planning for component replacement.

Maintainability: Worn contacts, damaged components, and accumulated contamination require periodic maintenance. Fixture design should facilitate inspection, cleaning, and component replacement without complete fixture disassembly.

Shielding Integration

EMC test fixtures often require shielding to isolate the DUT from ambient interference or to contain emissions during testing:

Enclosure design: Shielded enclosures around the test area provide isolation from factory EMI. The enclosure must accommodate product insertion while maintaining shield integrity during testing. Hinged lids, sliding doors, or specialized loading mechanisms address this challenge.

Penetration treatment: All penetrations through the shield (cables, pneumatics, material handling) require appropriate treatment to maintain shielding effectiveness. Filtered connectors, waveguide-beyond-cutoff tubes, and conductive gaskets prevent EMI leakage through penetrations.

Seam management: Shield seams at doors, access panels, and fixture interfaces require continuous electrical contact for effective shielding. Finger stock, conductive gaskets, or knife-edge contacts maintain conductivity across seams.

Absorber placement: For radiated emissions measurements, absorber material within the test enclosure reduces reflections that could affect measurement accuracy. The absorber type and placement depend on the frequency range of interest and the enclosure geometry.

Correlation Study Design

Establishing correlation between production and compliance tests requires carefully designed studies:

Sample selection: Correlation samples should span the range of EMC performance expected in production, including units near specification limits. Testing only typical units may not reveal correlation differences at critical performance levels.

Test sequence: The order of testing (production test first vs. compliance test first) can affect results due to handling, warm-up, or other effects. Randomizing test order or testing in both orders helps identify sequence-dependent effects.

Environmental factors: Temperature, humidity, and power line conditions may differ between production and compliance test facilities. Correlation studies should quantify these effects and either control them or include them in correlation factors.

Configuration control: Test configurations (cables, auxiliary equipment, operating modes) must be controlled during correlation studies to isolate the effects of the test systems themselves from configuration differences.

Uncertainty Analysis

Correlation factors have associated uncertainties that must be included in the overall measurement uncertainty budget:

Repeatability: Repeated measurements of the same unit on the same system show variation due to contact resistance, positioning, and equipment drift. This within-system repeatability contributes to correlation uncertainty.

Reproducibility: Measurements of the same unit on different systems or by different operators show additional variation. Between-system reproducibility is particularly important for correlation factors that must account for systematic differences between systems.

Regression uncertainty: When correlation factors are derived from regression analysis of paired measurements, the regression coefficients have confidence intervals that contribute to overall uncertainty.

Extrapolation risk: Correlation factors derived from a limited range of measurements may not apply accurately outside that range. Extrapolating to performance levels not included in the correlation study introduces additional uncertainty.

Correlation Maintenance

Correlation relationships can change over time due to equipment drift, environment changes, or process modifications. Ongoing monitoring maintains correlation validity:

Periodic verification: Regular testing of samples on both production and compliance systems verifies that the established correlation remains valid. The frequency of verification depends on the stability of the systems and the consequences of correlation drift.

Golden unit tracking: Stable reference units (golden units) tested periodically on production systems can detect system drift. Changes in golden unit results suggest system changes that may affect correlation.

Failure analysis: When products pass production testing but fail compliance testing (escapes), or when products fail production testing but pass compliance testing (false failures), investigation should determine whether correlation drift contributed to the discrepancy.

Change management: Changes to production test equipment, fixtures, or procedures require re-evaluation of correlation. Similarly, changes to compliance test facilities or procedures may affect the correlation relationship.

Test Time Analysis

Understanding the components of test cycle time identifies opportunities for optimization:

Setup time: Time required to load the DUT, establish connections, and configure the test system. Automated handling, quick-connect fixtures, and programmable configuration reduce setup time.

Measurement time: Time required to actually perform measurements. This depends on measurement bandwidth, averaging requirements, and the number of measurement points.

Processing time: Time for data analysis, limit comparison, and result reporting. Modern processors handle most analysis in negligible time, but complex algorithms or data transfers to remote systems can add delay.

Unload time: Time to disconnect the DUT and transfer it to subsequent operations. Automated handling and parallel processing can overlap unload with setup for the next unit.

Measurement Acceleration Techniques

Several techniques can reduce the time required for EMC measurements:

Reduced frequency points: Instead of continuous frequency sweeps, measuring only at specific frequencies known to be critical can dramatically reduce test time. These frequencies are identified through design analysis and correlation with compliance test results.

Wider resolution bandwidth: Emissions measurements with wider resolution bandwidth sweep faster but may miss narrow-band emissions. Understanding the product's emission spectrum enables selection of appropriate bandwidth for production testing.

Reduced averaging: Compliance measurements often use extensive averaging or maximum hold to capture worst-case emissions. Production tests may use less averaging with appropriately adjusted limits to account for the resulting variability.

Parallel testing: When the DUT has multiple ports or functions, testing them simultaneously rather than sequentially reduces total test time. This requires test equipment capable of multiple simultaneous measurements.

Equipment Selection for Throughput

Test equipment selection affects achievable cycle times:

Switching speed: Equipment switching between configurations, frequency bands, or measurement modes takes time. Fast-switching equipment minimizes dead time between measurements.

Settling time: After frequency changes or configuration switches, equipment must settle before accurate measurements can be made. Equipment with fast settling characteristics enables more measurements per unit time.

Data transfer: High-speed data interfaces between test equipment and control systems prevent data transfer from becoming a bottleneck. Modern equipment with USB3, Gigabit Ethernet, or PCI Express interfaces transfers data much faster than older interfaces.

Automation support: Equipment designed for automated testing includes features such as programmable state machines, trigger inputs, and rapid command response that facilitate integration into automated test systems.

Root Cause Analysis

Understanding why false failures occur is the first step toward reducing them:

Contact issues: Intermittent or high-resistance contacts can cause measurement errors that result in false failures. Contact analysis examines contact resistance trends, failure patterns, and correlation with contact maintenance schedules.

Environmental interference: Ambient EMI in the production environment can elevate measured emissions or cause spurious signals that appear as failures. Correlation with production activities or time of day may identify environmental causes.

Equipment malfunction: Test equipment drift, calibration problems, or intermittent faults can cause false failures. Equipment verification and maintenance records help identify equipment-related causes.

Limit errors: Incorrectly set limits, database errors, or software bugs can cause good products to fail. Review of limit derivation and verification of test software eliminate these causes.

Handling damage: Products may be damaged during test handling, causing failures that did not exist before testing. Handling analysis and damage inspection identify this cause.

Measurement Improvement Strategies

Technical improvements to the measurement process reduce measurement variability and associated false failures:

Contact improvement: Higher-quality contacts, better alignment, and proper contact maintenance reduce contact-related measurement variability. Contact verification measurements before production testing identify degraded contacts.

Shielding enhancement: Improved shielding of the test area reduces the effect of ambient interference on measurements. Shield integrity verification and regular maintenance ensure continued effectiveness.

Averaging optimization: Appropriate averaging reduces random measurement variation without excessive impact on cycle time. The optimal averaging level balances false failure reduction against throughput requirements.

Calibration improvement: More frequent calibration, better calibration procedures, or more stable equipment reduce drift-related false failures. The cost of improved calibration must be balanced against the cost of false failures.

Limit Optimization

Test limits directly affect false failure rates, and appropriate limit optimization can reduce false failures without increasing escape risk:

Statistical limit setting: Using production data to establish realistic limits based on demonstrated process capability avoids overly tight limits that cause unnecessary failures.

Guardbanding: Formal guardbanding methods account for measurement uncertainty when setting limits. Proper guardbanding ensures that the probability of false pass and false fail are appropriately balanced.

Multi-level limits: Using warning limits inside failure limits enables early detection of drift without failing units that are still compliant. Units exceeding warning limits receive additional scrutiny or trigger process investigation.

Limit verification: Periodic verification that production test limits correctly correspond to compliance limits catches limit errors before they cause extended false failures.

Data Collection Requirements

Defining what data to collect balances the value of information against storage and processing requirements:

Result data: At minimum, pass/fail results and limit comparisons must be recorded for each tested unit. This enables basic quality tracking and traceability.

Measurement data: Recording actual measured values (not just pass/fail) enables trend analysis and process capability studies. Measurement data supports correlation verification and test system qualification.

Configuration data: Recording test configuration (firmware versions, calibration dates, fixture identification, environmental conditions) supports troubleshooting and ensures traceability of test conditions.

Timing data: Test timestamps, cycle times, and equipment utilization data support efficiency analysis and capacity planning.

Diagnostic data: For failed units, additional diagnostic measurements help identify failure causes and support rework decisions.

Database Design

Production test databases must handle high data volumes while supporting the queries needed for analysis:

Schema design: A well-designed database schema organizes data efficiently and supports required queries. Separating measurement data, configuration data, and result data into related tables enables flexible analysis.

Indexing: Appropriate indexes on frequently queried fields (serial numbers, test dates, test stations) dramatically improve query performance on large databases.

Archiving: Strategies for archiving older data balance storage costs against data access needs. Online/offline tiering keeps recent data immediately accessible while retaining historical data for long-term analysis.

Backup and recovery: Production test data often has regulatory or contractual retention requirements. Backup procedures must ensure data preservation and recovery capability.

Analysis and Reporting

Raw test data becomes valuable through analysis and reporting:

Real-time monitoring: Dashboards displaying current test results, yields, and trend indicators enable immediate response to production issues. Automatic alerts notify responsible personnel when metrics exceed thresholds.

Historical analysis: Tools for analyzing historical data support process improvement, correlation studies, and capacity planning. Query capabilities should enable analysis by time period, product, test station, or other factors.

Standard reports: Regular reports summarizing yield, capability, and other metrics support management review and continuous improvement activities. Automated report generation ensures consistent and timely reporting.

Export capabilities: The ability to export data for analysis in spreadsheets, statistical packages, or other tools extends analytical capabilities beyond built-in functions.

Traceability Systems

Traceability links test results to individual products, enabling quality tracking through the product lifecycle:

Serial number tracking: Unique serial numbers on each product enable linking test results to specific units. The serial number system must be reliable and robust against data entry errors.

Genealogy tracking: For complex products, tracing components to their sources (lot codes, supplier, manufacturing date) supports root cause analysis when problems are discovered.

Process history: Recording the complete process history of each unit (operations performed, test results, rework history) provides a comprehensive quality record.

Customer linkage: Tracing products to their final customers enables targeted notification and support when issues are discovered after shipment.