Measurement data models

This reference organizes mpylab’s persisted data containers. It describes the supported read interface of a measurement history, not every internal attribute that Python happens to store when pickling a measurement class.

Basic structure

Measurement classes store several named datasets in one shared history object. The usual access path is:

history.<data_container>[description]

description is a freely chosen name that is unique within its container. Below it, workflows typically use numeric frequency keys in hertz and further keys for ports, tuner positions, points, detectors, or generator levels. The respective domain reference documents the exact order.

A description can collect several measurement acquisitions, including runs made after the setup changed. The compact acquisition reference on each bundled raw sample associates it with a timestamped configuration snapshot; Measurement history and output files defines this shared provenance structure.

Raw-data containers normally start with rawData. They hold instrument values, workflow metadata, and the provenance needed for repeated evaluation. processedData contains quantities derived from those raw values. An evaluation does not replace or delete the raw data.

Persisted containers

Measurement-class data containers

Measurement class	Containers	Content	Detailed reference
TEMCell	rawData_Verification, processedData_Verification	TEM-mode verification and power references	TEM verification data
TEMCell	rawData_e0y, processedData_e0y	measured GTEM field factor	TEM e0y data
TEMCell	rawData_Immunity, processedData_Immunity	exposure, reference selection, and EUT assessment	TEM immunity data
TEMCell	rawData_Emission, processedData_Emission	emission scan, detectors, and field correlation	TEM emission data
AmplifierTest	rawData, processedData	amplifier power, gain, and compression	Amplifier test data
MSC	rawData_MainCal, processedData_MainCal	empty-chamber and loading calibration	MSC calibration and autocorrelation data
MSC	rawData_EUTCal, processedData_EUTCal	EUT-specific chamber calibration	MSC calibration and autocorrelation data
MSC	rawData_AutoCorr, processedData_AutoCorr	tuner autocorrelation and independent positions	MSC calibration and autocorrelation data
MSC	rawData_Immunity, processedData_Immunity	exposure, thresholds, and EUT assessment	MSC immunity and emission data
MSC	rawData_Emission, processedData_Emission	received power and derived emission	MSC immunity and emission data

TEMCell additionally stores geometry, uniform_areas, verification_datasets, and reference_datasets. These attributes bind geometry and provenance to the measurement history and are relevant when interpreting verification and reference data.

Supported contract and internal state

The paths, types, units, and meanings described on these pages are supported. Later mpylab versions may add keys. Fields marked as optional can be absent from older pickles or another measurement mode; applications should read them with mapping.get(...).

Runtime objects such as open files, loggers, UI adapters, and instrument instances are not a data-reading interface. An attribute currently visible in __dict__ is not a stable contract unless this reference includes it.

Established keys are not silently renamed or assigned a new technical meaning. An unavoidable incompatible change requires a migration strategy and an explicit versioning decision. This documentation does not introduce a global schema version for the pickle history; individual interchange formats or nested documents can have their own schema_version fields.

Changing the data model

A change to a persisted attribute, key, index, type, unit, or meaning must update the German and English documentation in the same commit. Its virtual runtime contract test must be updated as well. New containers must appear in the table above.

See Maintaining the documentation for maintenance rules and check commands. Measurement history and output files documents pickle, autosave, and resume; Inspecting pickles with PExplorer explains interactive inspection of a trusted history.

Path-correction quantities keep their existing units and dictionary keys. Newly computed corrections retain SCUQ expressions, including shared-input and real/complex transfer dependencies, rather than independent evaluated inputs. Their pickle representation may include the internal transfer expression classes in mpylab.tools.ratio_conversion; readers need a version providing those classes. Existing saved numerical results are not rewritten. See Conditions, switching matrix, and path corrections for the corrected coherent power semantics and the limits of first-order uncertainty at cancellation.