High yields depend on reliable measurement

Paul Meyer explains how to improve test accuracy and throughput for optoelectronic components with specifically designed automated test instruments.

Paul Meyer explains how to improve test accuracy and throughput for optoelectronic components with specifically designed automated test instruments.

Automated testing is essential for controlling optoelectronic component costs and reliability.

If instruments do not deliver reliable measurements, investments in production test automation may be wasted.

Instrument reliability means more than long service life; it includes consistently high accuracy at high throughput rates, and data delivery that facilitates detection of yield problems.

Instruments that fail to supply these attributes can have a negative impact as significant as other automation equipment failures.

Component reliability parallels measurement reliability.

It is easy to detect a total failure in either one.

As a "brain-dead" instrument or electronic assembly doesn't do much, the solution is repair or replacement.

With early detection, the impact on cost usually is minimal.

Subtle failures and marginal performance are more difficult to detect.

Slow calibration drift is a subtle and insidious source of unreliable instrument readings.

It can involve basic instrument accuracy relative to a standard or calibration of electrical readings in terms of a physical quantity, such as optical power.

Instrument manufacturers usually publish thermal drift specifications, typically in terms of 90-day and one-year accuracies for a given temperature range.

If published specs are not available, it is best to ask the manufacturer to certify thermal drift.

Conversion of instrument electrical readings to other engineering units typically involves additional test system components, such as photodetectors, integrating spheres, fibre pigtail leads etc.

Calibration drift anywhere along the signal path can cause erroneous readings and process variations.

If this drift goes undiscovered, yield may be affected for days or weeks, depending on quality assurance methodology.

Calibration should be checked periodically.

As critical production facilities typically have multiple test stands and backup instruments, a simple calibration check is to cross-correlate readings.

This involves measurements on an identical group of DUTs using different test stands or instruments, or measurements on DUTs from the same production run taken over sufficient time to spot systematic yield differences.

In most plants, this should be an ongoing process where two production tools running identical processes pass work in progress (WIP) to two identical test tools.

This allows an engineer to correlate variations to a process or test tool.

It is difficult to detect random or intermittent measurement failures in the form of false "good" readings on bad parts or vice versa.

Passing bad parts may eventually increase field returns and warranty costs.

Rejecting good parts reduces yields.

Often, production test tolerance bands are purposely narrowed to avoid passing bad parts with instruments that have wide uncertainty bands.

Higher accuracy instruments may allow wider tolerance bands on production measurements, because higher precision provides higher levels of reliability and confidence.

This increases yields without increasing the "pass" rate for bad parts.

Many optoelectronic component test stands are plagued with frequent instrument lockups that require excessive operator interaction, create transposition error and lower throughput.

(Transposition error occurs as operators manually sort parts and erroneously record (transpose) data on those parts, resulting in false "defects").

Usually, lockups originate in the microprocessor or GPIB (general purpose interface bus) adapter card of an instrument that migrated from the research lab.

Such instruments are not designed for high throughput, 24/7 production operation.

Recovering a stalled instrument in an automated production environment is time-consuming and costly.

In addition to restarting the instrument, WIP stock must be rearranged to make sure all DUTs are completely screened.

Depending on test system complexity, this task can be complicated and prone to error.

Robust operation with minimal lockups depends on thoroughly testing instrument firmware in the user's type of production environment.

Nuisance failures tend to breed operator anxiety, which leads to throughput reductions as proper operation of test equipment is frequently checked.

A typical manufacturing response is to increase equipment investment, operator staffing, and maintenance technicians to sustain acceptable throughput.

This defeats the purpose of automation.

By combining novel measurement techniques with proven production test instruments, systems can be developed for high test accuracy, throughput, and component yields.

These solutions take advantage of automation features on instruments designed specifically for PC-controlled testing and can be scaled for different levels of automation.

Back to top