Doug,
you don't need to be so defensive, no criticism intended. I'm
very grateful that you've posted this that we can see and use for evaluations. I know I can request the additional files from you (I should have been clearer about that), and I probably will at some point as they will be very good input to algorithms. Just now I haven't, simply because there was no direct link to casually download them and I have not strictly needed more test material yet. I've put in a lot of hours in this analysis so far. When finished my algorithm will most likely outperform Capture One's current which will be good for everyone. Capture One isn't necessarily best at everything you know ;-). Lumariver HDR supports loading a raw file, apply LCC and export as a cooked raw DNG so one can further process in any raw converter with DNG support.
My analysis is based on what's actually there in the raw files, and compared to other digital backs and what is in their raw files. I see bad behavior as described in my analysis, and that it should primarily affect color fidelity very early on and demosaicing issues further out (mazing).
To me it's not acceptable that color crosstalk starts occuring with very small shifts, and I have
very strong indications that it's happening. Sure you can shoot a scene where color shift and desaturation is less important, but when I buy into a tech camera system I want very high image quality, and that includes color fidelity. I think many other customers also care about color fidelity, tech cameras is more than just wide angle images with sharp corners.
A test which would make it clearer is to shoot a color checker, first in the center, then further and further out and see how color reproduction change. If "usable image circle" is defined as the image circle where color is the same as in the center (ie no crosstalk) I'm quite sure it will be considerably smaller than the image circles you've stated so far. And that it's exactly the definition I've used in my analysis. You may think it's too strict, that's fine by me, but I would guess there are others that would prefer to use this definition too.
I think it's wise to be a bit pessimistic until these color issues have been thoroughly investigated. As it's impossible to see what's crosstalk and what's just pixel vignetting in raw analysis of a LCC shot one have to make a color checker test. I have very strong reasons to believe there is crosstalk early on though, as green channel separation and crosstalk happens about the same time on other digital backs. I also see color differences in the jpeg stitches that indicate the same.
If I was selling these high end systems I would for sure investigate this property. The thing is that a customer may very well make a mistake when buying into it. Detecting a color shift is not too easy especially if you don't know that you should look for it. Say if you rent it a couple of days and you don't know what to look for you may actually miss that the system can't reproduce colors right when shifted, and you buy into this back for doing your interior shots and only a month later or so you realize that color fidelity is not acceptable. I would then be disappointed if the dealer had not informed me.
You can't just shoot a test scene and "well this looks kind of good, then it must work". The sensor has been designed for accepting a certain max angle before crosstalk occurs, and the lenses have been designed to deliver a certain max angle. If I was buying into this system I would surely want to know from the seller how large image circles that are crosstalk free.
These library shot gives indications that crosstalk happens
early, but a color checker test or other method is required to give an appropriate answer. With your contacts in the industry you can probably get the design targets in terms of angles for both the sensor and lenses too, which is hard for me to get, as they're not available in the public data sheets.
I'm not here to try to trashtalk the IQ250. But I'm not here to raise it to the skies and try to get people buy it either, I'm just sharing my analysis based on my special expertise so potential buyers know what to look for.
Torger, I really don't agree with the pessimism of your analysis.
You say "none of the lenses" tested work well with the 250 and I just can't possibly understand how you come to that conclusion. Are there color issues on the outside of the image circle? Yes. But within the usable image circles the color and tonality and detail are really good.
Perhaps you should 1) request the raws of the 40 and 60 (you say "there are no raws of the 40/60" but in the article I clearly state we have them and are glad to provide them; we only opted not to post hard links to all 100 raws from the test out of concerns we'd overuse our storage account.
Also, as 99.9% of our customers are using capture one to work up their phase/tech files you might consider doing your analysis there, rather than inventing your own algorithms.
Of course you're free to come to your own conclusion - that's why I post the raw files and a test this complete.
