22.06.2011, 17:13 от Edmund Ronald wrote: I think white LED backlight displays are basically switched -on/off- with the duty cycle determining the brightness. Yes, in some cases the brightness is regulated by duty cycle. In some cases the diodes supplied by constant current (through an inductor) to reduce EMI and/or flickering. RGB LEDs can do on and off quite fast and in case of duty cycle regulation, there may be no spectrum changes. It's not the case with "white" LEDs. The changing of duty cycle or decreasing of DC through diodes can influence on the spectrum because of "phosphors" in white diodes. Has anyone looked at spectral drift here? There may be thermal effects involved. On Wed, Jun 22, 2011 at 2:11 PM, Nikolay Pokhilchenko <nikolay_po@xxxxxxx> wrote: 22.06.2011, 15:57 Edmund Ronald wrote: http://photofeedback.blogspot.com/2011/05/calibrating-macbook-pro-display.html Yes. But Juergen Lilienthere have proposed the spectrum correction for this case. The spectrum correction of white points may be useful, because the spectrum of light sources can changing significantly with brightness changing. For example, while dimming the CCFL. Please, excuse me for my English. On Wed, Jun 22, 2011 at 1:46 PM, Nikolay Pokhilchenko <nikolay_po@xxxxxxx> wrote: 22.06.2011, 14:50 от Juergen Lilien wrote: Many instruments especially spectrometers seem to have accuracy problems when measuring at low light levels. Couldn't it be beneficial, if we would first measure a display at its maximum brightness setting, subsequently set the brightness to the desired level, remeasure at least the new white point spectrum (I suppose the white point of the backlight unit will change when dimmed) and then try to compensate/recalculate the first measured data based on the new WP/brightness? I think it's a good idea, nor dumb. If one have spectrum data, he can just multiply this data by spectrum difference between white points, calculated after brightness changing. But in case of spectrum distribution change, this methode will work only for profiling, not for calibration. Could this strategy be beneficial at least for measuring the very low gray/brightness levels?I suppose it could. Graeme, it's a good idea. We can implement the modes of calibration and display reading at high brightness. I propose the next workflow which can be implemented in dispcal and dispread: 1. Calibrating desired white point and desired brightness. 2. Measuring and save desired white point spectrum. 3. Changing the hardware brightness of the display baclit to maximum. There is the check "Is the sensor of instrument saturated"? needed. If saturated, then operator should decrease brightness. 4. Measuring and save high brightness spectrum. 5. Calculating spectral "brightness correction coefficient" by devision the low brightness spectrum by high brightnes spectrums previously saved. 6. Doing the calibration or measurements with applying (multiplying by) the correction coefficient. That way the program can predict the spectrum of patch on screen at desired brightness, while the measurements have done at maximum brightness. I suppose this way can increase the tolerance of measurements and resulting calibration and profile. One reason for error I can see, is the display matrix transparency drift with temperature. At higher brightness the temperature of matrix will be higher. And if there is matrix temperature drift of color, there may be an error. But this error may be less than instrument error at low brightness.