Checkweigher Data

This chart illustrates a typical composite plot, resulting from the overlay of a computed curve on experimental results. In this experiment, 13 test packs were carefully made up at a range of weights spanning the likely error band of the automatic weighing belt. The range was symmetric about the intended set weight of 226g.

View example as VML

View example as SVG

The packs were then inserted into the production line (in random order) 50 times, and the packs rejected by the machine at each pass were noted. You can see from the chart that the 220g pack was always rejected, and the 229g and 232g packs were never rejected. However in the ‘zone of uncertainty’ caused by vibration of the weighing belts, there was a more or less consistent pattern, with more rejection of the lighter packs.

This can best be modelled with the ‘logistic’ curve which requires 2 parameters (the set weight and the uncertainty) to produce the illustrated function. This is exactly analagous to the Fermi-Dirac statistic followed by free electrons in a metal, where the energy band broadens with temperature in an identical way. A perfect checkweigher corresponds to a metal at absolute zero!

You will notice that the Y-axis caption renders perfectly on the thumbnail image, and also looks fine on the SVG example, but the VML renders it extremely badly. In general, you should avoid using rotated text if you intend to publish your charts in VML format.