Some forum members are purchasing strong ND filters (Hoya NDx400, B+W110,...). I was wondering of a test we could do to find out the response of those filters to each part of the visible spectrum.
The experiment I am planning is not scientifical at all, and will not provide response as a function of wavelength, but could provide response as a function of Hue, which can be similar to the former and can be useful anyway to better know how out filter works in terms of colour casts introduced by the filter.
I have designed a test chart with all the Hue values (0..360ยบ cycle) at different saturation levels (100%, 75%, 50%, 25%). I added a neutral gray scale with 5 values in case I want to do some considerations including white balance and so forth.
My intention is to ask for 2 RAW files obtained shooting the chart displayed in the monitor
without and
with the filter on the lens. Exposure time will be corrected up in the second shot to obtain a reasonable high degree of exposure. I am then planning to compare the RAW levels obtained in both shots for each channel obtaining an attenuation coefficient in EV for each Hue and camera's RGB channel. Will find out if saturation as defined in the HSV colour space can affect the result.
Assuming the results will be more qualitative than quantitative. We are mixing up different uncalibrated systems than make impossible to properly obtain a frequency response:
1. Chart is not pure wavelenghts (should be LASER light), so each Hue is participated of many wavelenghts in the visible spectrum.
2. Monitor will not be calibrated so neutral gray patches are not (this could be fixed by white balancing the shots, but the WB needed with or without filter would be different so results could make no sense).
3. Our sensor has different sensitivities and captures a wide band of the spectrum so each channel will be influenced by many variables.
Even with all those aproximations and variables, do you think it makes some sense to do this test?
This is the chart I designed:
Here in 1024x1024:
ND test chartSomething like this is what I wanto to achieve:
But it will be for each separate RGB channel. Perhaps the results can be combined to get an aproximation to the real luminance response of the filter.
In the end, if it's true that Hue is a good variable to characterise the filter response, some HSV Hue curve (not an RGB curve):
Hue_OUT = f(Hue_IN)
could be apllied to the RAW data to "calibrate" each filter eliminating its particular colour casts. If saturation also happens to affect the response (which is expected since it affects the wavelengths present in the tone), the Hue curve would be 2-dimensional, depending on 2 variables:
Hue_OUT = f(Hue_IN, Sat_IN)
