Dataset Preview — Reading Calibration & Quality

The Calibration & Quality tab of the Dataset Preview gathers information about the calibration and the quality checks published together with the scientific curves of the Curve Digitalizer. It lets you check source identity, units, accepted calibrations, data consistency and raster diagnostics without modifying the Dataset.

The example uses a synthetic graph with three manually digitized curves, each with eight points, on a raster of 1280 × 900 pixels. The published Dataset contains 24 rows and a single accepted calibration. The quality indicators describe software checks and raster characteristics: they are not automatically equivalent to a metrological evaluation of accuracy or uncertainty.

1. Two Tools with Different Tasks

The Curve Digitalizer Calibration tab serves to set up and accept the transformation from pixel coordinates to scientific coordinates. Calibration & Quality, instead, belongs to the Dataset Preview: it shows the results and the information recorded by the publication.

There is no need to return to the Viewer to read these metadata. Do not use the Preview tab as if it were a command to correct a calibration: the information shown is intended for consultation.

2. Completing the Calibration in the Curve Digitalizer

The example starts from the definition of the axes Elapsed time (s) and Detector response (mV), both linear, with X from 0 to 10 s and Y from −20 to 80 mV. The calibration requires the four references BL, BR, TL and TR and is accepted in the Viewer.

Accepted four-point calibration with Elapsed time and Detector response axes
Figure 1. Four-point calibration accepted on the synthetic raster, with scientific ranges and units visible.

Acceptance describes the completion of the procedure in the software. It does not in itself constitute an independent verification with external experimental references.

3. Preparing Three Curves with Their Points

Three independent curves were created, Bias -2 V, Bias -1 V and Bias +0 V, with eight manual points per curve. The two further lines in the raster graph remained simple elements of the image.

Curves tab with three series and 24 points
Figure 2. The Viewer’s Curves tab with the three series and 24 original points ready for publishing.

Knowing the number of curves and points is important in order to compare it with the counts displayed later in the Quality tab.

4. Choosing a Realization to Publish

In the Publish tab the Viewer presents the accepted calibration realization and the planned scientific output. The example uses separate, that is, a Dataset corresponding to a single selected realization.

Publish plan with one accepted calibration in separate mode
Figure 3. The Publish plan with an accepted calibration and the identity of the output numerical Dataset.

This choice will be reported in the metadata as Calibration mode: separate, distinct from other possible publishing modes.

5. Publishing the Dataset Before Consulting Quality

Press Publish selected calibrations and check that the publication has been registered. The Quality information is not an analysis to be built by reading the image directly: it accompanies the canonical Dataset produced by the Viewer.

Publish confirmation of the 24-point Dataset
Figure 4. Successful publication of the numerical Dataset containing the 24 points of the three curves.

The published output does not replace the raster source or the editable document in the Viewer.

6. Opening the Correct Dataset in the Framework

In the main window choose Dataset → Preview and select the numerical Dataset, not the previous Image Dataset. In the example the result has the table.generic contract and presents the framework’s scientific consultation tabs.

Numerical Dataset selected in Dataset Preview
Figure 5. The numerical Dataset selected in the GX-TXT browser before opening the specialized Quality tab.

The identifier displayed belongs to the exercise Session. In a different project always use the actual identity of the published Dataset.

7. Opening Calibration & Quality

In the row of tabs choose Calibration & Quality. The panel presents a text consultation with the PREVIEW METADATA section and various field-value pairs. It is not the digitizing Canvas and contains no buttons to accept or move points.

Calibration & Quality tab with PREVIEW METADATA
Figure 6. The Calibration & Quality tab of the Dataset just published, in its initial position.

To read all the information, use the panel’s vertical scroll bar. Some diagnostics are longer than the visible portion.

8. Reading Dataset, Contract and Source Dataset

The first fields indicate Dataset, Contract and Source Dataset. They distinguish the numerical result, the type of representation and the image from which the point positions were derived.

Dataset, Contract and Source Dataset fields
Figure 7. Original detail of the header with the Dataset identity and the main publication fields.

The scientific Dataset of the example is distinct from the source raster. Their association is kept in the provenance, not in the position of the windows on the screen.

9. Understanding Source SHA-256

Source SHA-256 is the digital fingerprint recorded for the original image. The value serves to identify the source exactly, even when two files have the same name or the same dimensions.

In the documented case the source check returns verified. The identity of the image must not be replaced with a visually similar preview: the colour of a line is not proof of provenance.

The Source SHA-256 field shows the digital fingerprint associated with the image used for digitization. When the Dataset is consulted later, this reference helps distinguish the exact origin of the pixels from other images with similar names or appearance.

Source verification concerns the consistency of the raster identity. It does not prove the correctness of the units, the accuracy of the calibration or the absence of experimental uncertainties. For this reason the Source SHA-256 field must be interpreted together with the other states of the Quality report.

10. Calibration Class and Calibration Mode

The panel shows Calibration class: affine and Calibration mode: separate. The first field describes the declared class of the transformation, the second the mode in which the realization was published.

Do not confuse these values with the characteristics of the graph: a declared transformation and a successful publication do not provide an estimate of the physical uncertainty of the measurement.

11. Calibration Realizations Does Not Mean Experimental Replicates

Calibration realizations: 1 indicates a single published calibration realization. It is the number of configurations available in the result, not the number of repeated instrumental trials.

If only one realization has been registered, a statistical dispersion between calibrations cannot be derived automatically. Do not assign zero uncertainty to a dispersion that cannot be calculated.

12. Recognizing the Scientific Variables and Units

The Scientific X Variable and Scientific Y Variable fields identify the columns of the Dataset. Their labels are Elapsed time, in s, and Detector response, in mV.

The recorded units come from the calibration and publishing work. They do not represent information inferred from the Plot magnification or from the raster dimensions.

13. Family Quantity, Family Unit and Grouping Identity

In the example Family quantity and Family unit are Not declared because the three curves were published as independent series and not as a parameterized numerical family. This is not an error.

Grouping identity: curve_key indicates the column used to distinguish the curves stably. The readable names of the series help the user, but the canonical identifiers remain the reference for associating their points.

14. Quality Status and Quality Issues

The publication shows Quality status: ok and Quality issues: 0. This means that the report did not record problems in the check categories provided for this Dataset.

These two fields do not mean that the image was measured with known accuracy. The diagnostics of the acquisition, the geometry and the raster must be interpreted within their declared scope.

15. Curve Count, Original Raster Points and Published Rows

The values are Curve count: 3, Original raster points: 24 and Published rows: 24. They are consistent with the three eight-point curves created in the Viewer.

This agreement documents that the original points were represented in the publication without an unexpected change in the count. It is not a quantitative verification of the error in abscissa or ordinate.

16. Reading Calibration 1 Status and Its Axes

The tab describes Calibration 1 status as Measurement 1 / accepted / affine and reports the internal identifier of the realization. Its axes are X 0.0 to 10.0 s; Y -20.0 to 80.0 mV.

The axis labels and the fact that the realization is accepted can be checked directly by comparing the tab with the work carried out in the Viewer.

17. The Diagnostics Start from the Origin of the Points

Scrolling the panel, the Quality detail entries appear. The first describes Point origins: in the example 24 manual were recorded, that is, 24 manually acquired points.

Quality detail entries with point origins and segment diagnostics
Figure 8. Quality diagnostics scrolled down to the per-curve details and the geometric parameters of the raster.

The manual or assisted origin documents the acquisition procedure. It does not, on its own, constitute an estimate of the metrological quality of the tracing.

18. Segment Report Coverage and Number of Segments

A curve with eight consecutive points normally produces seven segments between adjacent pairs. Three curves of this type therefore generate 21 segments; the report indicates 21 of 21; truncated=No.

The value describes the coverage of the check on the published segments. It must not be interpreted as 21 new independent experimental observations.

19. Local Image Evidence: Evidence from the Raster

The report shows Local image evidence available: Yes. For the example, 21 segments were examined with the method indicated as local_center_vs_perpendicular_flanks, using a contrast threshold of 12.0 in 8-bit intensity units.

This is a check on the local structure of the image: grids, crossings, labels and contrast can mimic a curve. Contrast evidence is not equivalent to physical accuracy.

20. Distinguishing Pixel Spacing from Physical Uncertainty

For each curve, minimum, median and maximum distance values between consecutive points are available, expressed in px. The report indicates, for example, for the Bias -2 V curve a minimum distance of about 105.04 px and a maximum of about 118.62 px.

These distances describe the geometric sampling of the points in the raster. They are not standard uncertainties in seconds or millivolts and cannot be used directly as error bars.

21. Local Visible Fraction and the Limits of Contrast

Another indicator is local visible fraction, calculated on the image samples used for the segments. In the example some fractions are below 1 even though no general quality problem was reported.

The fraction does not measure the probability that the curve is scientifically correct. It serves to describe how much of the local signal shows the image evidence expected by the check.

22. Valid Geometry and Conditioning

Towards the bottom of the panel Calibration Measurement 1 geometry valid=Yes appears, with a homography condition value of around 5100.35. The condition depends on the scale of the coordinates and is a numerical-geometric indicator.

Geometry validity, homography condition and reprojection RMS details
Figure 9. Original crop of the geometric details and of the warnings on the calibration verification.

Do not use the condition number as a direct evaluation of the precision of the instrument or of the absolute goodness of the calibration.

23. Control Reprojection RMS Is Not Independent Accuracy

The GUI reports a control reprojection RMS of about 1.62 × 10⁻¹⁶ for the same four references used in the transformation. The value is very small because it refers to the reprojection of the controls used to determine the calibration.

It is not a verification with independent points and must not be interpreted as zero physical uncertainty. An almost zero numerical residual can coexist with errors in the choice of the original references.

24. Independent Verification: Not Recorded

The report indicates independent verification: available=false; status=not_recorded. Results of a verification distinct from the four calibration controls are therefore not available.

The absence of verification must be reported as such: it does not authorize declaring that the experimental error is zero. Similarly, a single realization does not automatically provide a dispersion between calibrations.

25. The Three Final States Are Not Equivalent

At the end of Calibration & Quality you can read Source verification: verified, Accuracy status: not_assessed and Uncertainty status: not_declared.

[IMAGE 10] — 12_native_Quality_accuracy_uncertainty_status_bottom.png

Figure 10. Bottom of the tab with information on the geometry, independent verification not recorded and final states.

The verified identity of the source confirms the consistency of the declared link within the scope of the available checks. It does not certify the physical measurement; accuracy remains not assessed and an undeclared uncertainty cannot be replaced with zero.

26. Ordinary Quality and Calibration & Quality

In addition to the additional Calibration & Quality tab, the CORE Preview presents a Quality tab. The latter organizes the same published information in rows with the columns Diagnostic, Value, Unit and Meaning.

Ordinary Quality tab with Diagnostic, Value, Unit and Meaning columns
Figure 11. The ordinary Quality tab shows a readable table with the publication diagnostics.

The additional view makes compact reading of the calibration metadata easier; the Quality table helps consult the meaning associated with the individual indicators.

27. The Unit and Meaning Columns Avoid Wrong Interpretations

In the Quality tab, the Unit column makes it possible to distinguish counts from distances in pixels, while Meaning explains the scope of the value. The presence of a numerical value does not imply that it is a measured physical quantity.

Detail of the Unit and Meaning columns
Figure 12. Original detail of the diagnostic table with declared units and meanings.

The interface distinguishes the description of the check method from its result: read both before reporting a conclusion in the scientific documentation.

28. Selecting Physical Accuracy

In the Quality table click the Physical accuracy row. The lower Selected diagnostic panel shows the value not_assessed and specifies that the Quality contract does not automatically infer a measured accuracy.

Physical accuracy row selected with not_assessed value
Figure 13. Selection of the Physical accuracy entry in the ordinary table; the panel explains why it is not an accuracy measurement.

Selecting a row does not change the Dataset. It is an operation for consulting details already recorded.

29. Selecting Uncertainty Declaration

Click Uncertainty declaration. The value is not_declared: it means that an uncertainty declaration is not available in the publication.

Uncertainty declaration row selected with not_declared value
Figure 14. The interface explains that an undeclared uncertainty does not correspond to an uncertainty equal to zero.
Selected diagnostic panel for the uncertainty declaration
Figure 15. Detail of the Selected diagnostic panel with the explicit warning about the missing uncertainty.

If physical uncertainties are needed, they must be provided through a model or a documented procedure and published consistently. Do not derive them arbitrarily from the line width or from the control RMS alone.

30. Verifying the Source Provenance

The Provider-reported source integrity row shows verified. The description specifies that the provider’s result must not be confused, on its own, with an independent and complete verification of the bytes by the CORE.

Provider-reported source integrity row selected
Figure 16. The selected source integrity diagnostic shows the value and its limitation.

The SHA-256 recorded for the original raster must be considered together with the information on the Session and on the provenance of the publication.

31. Reading Reported Issues

The Reported issues row is 0 count. It expresses the number of problems published by the diagnostics system for this Dataset and for the rules applied.

Reported issues row with 0 count
Figure 17. The selected Reported issues row indicates zero diagnostic problems, not a physical error equal to zero.

A correct description in the report is “no diagnostic problem reported”, not “experimentally perfect curve”.

32. Returning to the Calibration & Quality Tab

After consulting the ordinary table you can return to the Calibration & Quality tab to find the calibration summary and the entire text list again. Switching between the two tabs does not require a new publication.

The scrolling, navigation and diagnostic selection operations modified neither the raster source nor the canonical content of the published Dataset.

Returning from Quality to Calibration & Quality you find the same panel already illustrated at the start: there is no need to reproduce a second screenshot of it. Check that the Dataset identifier has not changed and that the calibration information always refers to the same publication.

Switching between the two views does not repeat Publish, does not generate a new report and does not automatically correct scientific data. The three states Source verification, Accuracy status and Uncertainty status must be reread separately.

33. The Full Quality Report and the Summary Information

The CORE table also exposes the buttons Inspect full published report… and Copy full published report. They are commands to examine or copy the published diagnostic report in depth, not to change the accepted calibration.

The publication in the case illustrated contains a Quality Report v2, with data on curves, segments, local evidence and calibration. The full report is richer than the summary visible at the top of the tab.

34. Checking Consistency Before Using the Data

Before citing a result, check the selected Dataset, the source ID, the X/Y units, the number of realizations and the number of points. Check Quality status, Source verification, Accuracy status and Uncertainty status separately.

A value can be available, consistent and published without its physical accuracy having been assessed. The meaning of each state must remain distinct.

35. Summary of the Scientific Reading

The operating sequence is: Dataset → Preview → select the numerical Dataset → Calibration & Quality. Scroll through the information down to the calibration details and the three final states; to interpret the individual indicators also use the CORE Quality tab.

In the example there are one accepted calibration, three curves, 24 manual points, 21 segments and no diagnostic report. The source is verified, but physical accuracy and uncertainty are not declared as assessed. It is this distinction that makes reading the Quality report useful for scientific work.