Semi E151

SEMI E151 defines a taxonomy and guideline for understanding data quality in fab data collection — classifying the specific ways a reported value can be wrong (sensor drift, dropped samples, timestamp misalignment, stale cached values, unit-conversion errors) so a host analyzing equipment data knows what kind of quality problem it's dealing with, rather than treating "bad data" as one undifferentiated category. Every standard this family has built so far assumes the data arriving is correct — E136 verifies power-sensor accuracy at the point of measurement, E148 ensures timestamps share a common clock, E142 confirms a mapping result reflects reality — but none of them, individually or together, give a host a systematic way to diagnose a data-quality problem once one slips through, because "the number is wrong" has many distinct root causes that each require a different fix.

A reported value looks suspicious or inconsistent with expectations
Individual standards (E136, E148, E142) each guard against one specific error type
SEMI E151 — data-quality taxonomy and guidelineClassifies the distinct root causes behind a bad value
Named categories: sensor drift, dropped samples, timestamp misalignment, staleness
Each category points toward a different diagnostic path and fix
Guidance for flagging and reporting which category applies
A host can act on stale data differently than miscalibrated data
Analytics and APC systems can diagnose and route data-quality issues correctly
Not just discard suspect data, but understand why it's suspect

The reason data quality needed its own taxonomy standard, distinct from every individual accuracy or timestamp standard already built, is that those standards each prevent one specific failure mode at its source, but a host consuming the resulting data stream still needs a way to recognize and classify a quality problem that gets through anyway — a value that's technically well-formed and on-time but subtly wrong because of sensor drift looks completely different, diagnostically, from one that's correct but simply stale because a collection plan lagged, and conflating the two leads to the wrong corrective action. E151 gives a host that missing diagnostic vocabulary: instead of a binary good/bad judgment on incoming data, it provides named categories for the distinct ways data quality actually degrades, so an analytics pipeline or APC system can respond appropriately to the specific problem rather than treating every anomaly identically.

One Suspicious Value, Four Possible Diagnoses "bad data" is not one category — each root cause needs a different fix SVID 2004 = 148.3°C looks wrong — why? Sensor drift consistent offset, trending over weeks fix: recalibrate Dropped samples gaps in the expected sample rate fix: check E134 collection plan Timestamp misalignment value correct, time tag is wrong fix: check E148 clock sync Stale value cached, hasn't actually refreshed fix: check refresh interval Without E151's taxonomy: one undifferentiated "bad data" flag Same symptom, four different root causes — guessing wrong wastes the fix on the wrong system every time

E151 is what gives a host software consuming fab equipment data the missing diagnostic layer every individual accuracy and timing standard assumed it wouldn't need — by naming the distinct root causes behind a bad value instead of treating anomalies as one undifferentiated category, it lets an analytics pipeline or APC system route a data-quality problem to the fix that actually addresses it, rather than discarding or flagging suspect data without understanding why it went wrong in the first place.

Take SEMI E151 further

Ask the copilot about this term, or have our engineers assess it against your process.

Ask CFSGPT about this → Request a technical assessment → Open the CFS app