DP1s Sensor Response Curves (from a Prism)
Being a little underwhelmed with the DP1s colour, a few weeks ago a bought a prism and took a photo of the rainbow it produced with the afternoon sun shining through it onto a sheet of A4 paper stuck to a curtain:
I was actually fairly happy with the colour reproduction of the rainbow. Although in real life the cyan was more vivid.
I decided to graph the RAW values from the image. I also scaled the wavelength axis because blue light bends more than red through the prism (I used the Sellmeier equation). This is the graph I got:
The wavelength values shouldn't be taken as accurate I just played with them until the graph 'looked right' (on the x axis that is, the relative amplitudes of each channel are not changed).
The graph looks fairly similar to that in Foveon's paper:
However, one thing I noticed is that the red and green channels in my DP1s seem to be fairly highly correlated. That is there isn't much difference in the shape of the curves to detect how red something is.
In my own raw software I've been working on I've found that I get a fair bit of red grain in the image and I wonder whether it has something to do with very little chroma information for green-red?
Today it occurred to me that the red response for the DP1s might be cause by a fairly strong IR cut filter. Looking at this forum post of two photos comparing the DP1 and DP1s IR cut filters, the DP1s version appears more green which may indicate it is cutting out more of the red:
So my current thoughts are to remove the DP1s IR cut filter and to replace it with something like the B+W 486 UV-IR cut filter which has a fairly steep cut off at about 700nm:
In the Foveon paper they also show the response curves without the IR cut filter and it appears that the red channel increases until about 650nm, after that all three curves seem fairly highly correlated so it would seem to me that a fairly steep cut off would be better than a gradual one.
Any thoughts on this?