Gerhard Fuernkranz skrev 2012-02-11 17:59:
Am 08.02.2012 17:49, schrieb Florian Höch:Am 07.02.2012 23:36, schrieb János, Tóth F.:I understand a ccss file won't forcefully align the two meters but now I should ask myself which one is more accurate. Should I create matrix corrections instead or accept the i1d3 results?Good question. In my case a CCMX brings the i1d3 readings pretty close to those of the i1 Pro (see above).The aim of ccmx is clearly to make a tri-stimulus colorimeter agree with a given reference instrument, on a particular display. So if your goal is the same, namely to make two meters agree (regardless of absolute accuracy), and if you have access to both meters so that you can measure the same display with both meters, then use ccmx.Ccss files don't have the aim to calibrate an i1d3 against a reference meter. Readings obtained with the i1d3+ccss still rely on the factory calibration of the instrument. The ccss file just describes spectral characteristics of a particular display. In fact, the aim behind the scenes of a ccss file is basically the estimation/reconstruction of the spectra emitted by the display, from the tri-chromatic readings obtained from the i1d3's sensors. These estimated spectra can then be converted to XYZ in the usual way by applying e.g. the 2° or 10° standard observer CMFs.If only XYZ numbers need to be reported, then the spectral reconstruction and the subsequent XYZ integration can be still combined into a simple 3x3 matrix (derived from the ccss and from the sensor sensitivities) which is then applied to the (linearized) readings from the sensors in order to yield the corresponding XYZ numbers. But theoretically, also the estimated spectral readings could be reported by the i1d3 driver, given a ccss file which is representative for the display (or spectral readings could be even estimated from XYZ readings, by applying a ccss file to the XYZ readings).The estimation/reconstruction of the spectra is still an approximation and the accuracy is of course limited. For instance, a linear estimation model needs to assume a perfectly additive display, whose emission spectra can be represented as linear combination of three basis spectra (which is not granted in practice for many displays). The resolution, accuracy and representativity of the ccss for the particular display play a role too, and so does the accuracy of the sensor sensitivities measured by the manufacturer and stored in the instrument.The good thing is however, that the accuracy of the spectral reconstruction becomes less and less important for the accuracy of reported XYZ numbers, the closer the sensitivities of the instrument's sensor approach the standard observer CMFs (or linear combinations thereof) - and the i1d3 sensors are said to be pretty close.Best Regards, Gerhard
Very enlightening and thorough description. Thanks! /Kristian
