If the results files you are opening were created by a system running HardingFPA-X Analyser Version 2.0.0 or above, the analysis now features new analysis algorithms, which are better tuned to High Definition and File-based work.
They are better suited to subtle changes in the image data, and provide much closer results when testing the same material repurposed either into a different video format, or encoded with a different codec. The main differences between the legacy algorithms and the new ones are detailed below:
Differences Between v2.5 and v3.3
Different Graph Scaling
The HardingFPA generates risk values using the same range as its predecessor (i.e. 0 to 3.4) but displays the graphical data using a revised vertical scale. This modified scale allocates much more vertical space for risk trace warnings and diagnostic trace steps but only displays risk traces up to the value of 3.0. Risk traces values from 3.1 to 3.4 are still logged as part of the results files but are graphically displayed capped at 3.0.
(left) Version 2 and (right) Version 3 Graphing
The Squiggle
The HardingFPA gives enhanced visual diagnostics when an incoming transition coincides with an outgoing transition from one exactly second earlier. The new version 3 algorithms insert a squiggle (see below) to indicate when the diagnostic trace has simultaneously gained and lost a transition over the most recent second between video frames.
Analysis Results
The HardingFPA gives results which are broadly similar to those generated by version 2.5. The figure below shows the results of both versions when analysing the same video input under the same guidelines:
However, the results between the two versions will not be identical. The HardingFPA will, in general, be more lenient to complex, rapid motion:
... but more strict to examples of powerful, localised flashing:
Most importantly of all, the HardingFPA has been designed to be as format-agnostic as possible. Changes in file formats or codecs will alter the underlying video data even if these changes are not visually apparent. Here, the same video has been encoded at the same resolution using two different codecs. The absolute differences between the two images are shown in the third image as deviations from mid grey.
The new version 3 algorithms generate highly consistent results from different image resolutions and frame rates. Here, the same movie has been analysed in 720x576i50 and 640x480i60 formats with highly consistent results. The only noticeable difference is the different horizontal graph scales as a result of the different frame rates.
The graph below shows three sets of luminance flash risk results of the same movie analysed in SD-576i50, HD-720p50 and HD-1080i50 formats. The results are sufficiently similar it is difficult to see that the graph contains three separate traces: