Image Professor tip on color sky issues with tech cams

[Wayne Fox]

I assume most of you using Capture One get the Image Quality Professors emails and maybe check out his posts, but perhaps some of you are like me and just sort of ignore them since I’ve been doing it so long I get stuck in a rut.

But this one was interesting and seemed useful to those of us trying to get good results out of our tech cameras, so I thought I’d post it here.

3 easy steps fixing skys with the skin tone tool

 

[Paul2660]

Thanks for posting. I am glad to see the Phase admits that the LCC alone does not fix a sky with movements.

I have been doing a similar method of this for a while now.

I create a ND layer from the far corner towards the middle then make a selection in the upper corner with the color selection tool and then tweak the hue and sat that way. The only issue is that you can't put another layer over the first and make another color selection as it will get an error in the selection. You can however still re-tweak the hue and sat some more.

I will try this new method out as it looks to be possibly a better way.

Nothing harder to correct than a solid blue with a tech camera and movements especially with wides.

Paul C

 

[torger]

Really nice work-around for a very real problem. It's short-term though, many of the current tech cam users can probably live with it, but when even the LCC shot (which "normal" camera users already dislike) doesn't fix the cast I think it just becomes too much for new users.

Please Sony -- make us a sensor with light shields between the pixels, just like the Kodaks had.

 

[tjv]

Am I right in thinking the example picture in the article is take with an IQ100, or at least a cmos sensor? I've never seen anything like this amount of crazy colour before or after applying an LCC with my Credo 60.

 

[torger]

Am I right in thinking the example picture in the article is take with an IQ100, or at least a cmos sensor? I've never seen anything like this amount of crazy colour before or after applying an LCC with my Credo 60.

It's CMOS. You don't get this issue with the larger pixel CCDs, at least not at this magnitude.

 

[Paul2660]

In their example the red cast is actually pretty minor IMO. With my IQ260 or 160 I could easily get the to the same amount of red cast with the SK35 at 8mm of shift or the 40mm HR-W with 15mm of shift.

Paul C

 

[MGrayson]

After an LCC? I've never seen a corrected image that bad from an IQ160. I've used an IQ140 and 35XL and shifted past the usable image circle, but the problem has never been post-LCC color.

--Matt

 

[Paul2660]

After the LCC no. But in the first straight out the camera example before the LCC was applied, that amount of red cast is pretty common at least for me.

And after the LCC is applied I almost always will need to tweak a solid blue sky like the one in their example. Especially towards the edge of the shifted images.

Paul C

 

[torger]

It's CMOS. You don't get this issue with the larger pixel CCDs, at least not at this magnitude.

To make my answer a bit clearer. If there's a residual color cast after LCC removal the reason is almost always crosstalk. All Rodenstock Digaron wide angles have been designed to be compatible with the Dalsa 6um sensor. As retrofocus design is a drawback when making super-sharp low-distortion wides they've kept them as little retrofocus as possible to not introduce any significant crosstalk on that Dalsa 6um. There's still issues with tiling and microlens ripple which I think is something that Rodenstock missed in the design, but fortunately that can be cleaned up automatically quite well.

Anyway, with the as-little-retrofocus-as-possible principle this means that even the only slightly smaller pixel on the 80MP Dalsa (5.2um) get some issues with the Digarons, but not as bad as with the Sony CMOS.

The Schneider Digitar wides (and longer lenses too like the SK60XL) are symmetrical and doesn't care about what angular response the sensor's pixels have. Schneider obviously hoped for that Kodak's design with light shielded pixels would survive/return in the future but they were wrong (so far). So indeed they don't work too well even with the Dalsa CCDs, you can get large amounts of crosstalk for the widest and thus the same effect as CMOS on Rodenstock Digarons. As far as I know Dalsa has never had light shields either, and indeed I could get visible crosstalk artifacts even with my previous back a 33 megapixel Aptus 75 on the SK35XL for extreme portrait shifts, but the problem gets worse the smaller the pixel are.

The foundation of tech wide superiority is the ability to allow small widest aperture and little or no retrofocus. The more retrofocus the lenses become the more similar they will become in performance to normal SLR lenses. I suspect that it's more or less meaningless to make a new tech wide lens series which is even more retrofocus to support Sony CMOS in full, as the design would be too close to an SLR lens.

So either Sony needs to come up with light shielded pixels or other technology to reduce the crosstalk, or even the Rodies will grow obsolete and fade away just as the SK Digitar series have. There will still be a few fans left for ages though, like myself

 

[Paul2660]

It's really a matter of preference I guess, I have thousands of shots with the IQ160, and IQ260, and I almost always shift, the "before" image from the 100MP on the site to me doesn't look as bad as what I saw with the 50MP when shifted as past 10mm the shift is almost pure red, albeit the LCC cleans it up very well.

For my work, unless it's a pure blue solid (clouds can help hide the problems a lot as you eye won't see it as quickly) I just don't see that big an issue on the CMOS chips on large shifts. I was very pleasantly surprised at just how much work on returning the correct colors and saturation C1 did on the IQ150 shots I have. Yes if you zoom in past 100% you might start to see some artifacts (crosstalk) but for my work the results are fine. Plus as I stated before, the amount of magenta shift in the "before" shot on the blog post, looks like many thousands of IQ160 and IQ260 shots with Rodenstock shifts, and trust me, the LCC won't totally clean up those shifts either and the requirements to do something like in my first post or the blog post were always there, at least for me.

For me it's simple, the advantages of CMOS for my work far out weight the possible minor problems that crosstalk may introduce on shifts.

Who knows that the future will hold from Sony, but a dedicated chip for Tech? I don't see that in the cards.

Paul C

 

[voidshatter]

Who knows that the future will hold from Sony, but a dedicated chip for Tech? I don't see that in the cards.

Paul C

It's already known that BSI and stack will be the future. Staff from Sony's sensor department has confirmed with me. They already have some prototypes.

 

[Paul2660]

It's already known that BSI and stack will be the future. Staff from Sony's sensor department has confirmed with me. They already have some prototypes.

Personally, not planning to wait, as I have not been all that impressed with the BSI on the A7RII, for both shifts and higher ISO, the later especially. I am sure we will see something new that's what marketing is all about. And more than likely as the current 100MP is backordered way out to late this year at least for the Phase One side, not sure why Sony/Phase is really going to push anything forwards.

Phase left off video/sound on the current chip, fully expect that to be implemented sometime in the future.

Instead of BSI, I would have just liked to see 60 to 75MP with less pixel pitch full frame.

Paul C

 

[torger]

Angular response is an issue in sensor design that is considered by current manufacturers, just not on the extreme scale we see tech wides require.

Smaller pixel pitch makes pixels deeper (relatively speaking) which narrows down angular response. BSI makes pixels shallower, but what will that be used for? Maybe to make pixels even smaller while maintaining the angular response of today.

I had hoped that BSI would greatly increase angular response as a side effect, but the A7r-II results I've heard about (have not got test files personally) has not made me hopeful. I am a bit surprised that it haven't shown better results though, I was expecting more. It would be interesting to know why it fails to drastically increase angular response. Or maybe the angular response is good, but the microlenses introduce crosstalk over some critical angle? There's still no light shielding there.

I don't know if we really need to drop down to 60MP to get some decent angular response again, what probably is needed is that Sony's sensor designers get a different design target. I don't think their design target is "make as wide angular response as possible with current technology", but rather like "there's no need to have wider angular response than X degrees" and they design the pixel for that, which probably means a certain shape of the microlenses, no need for light shields etc.

I'm quite confident that they already have the toolbox required to do something better than Kodak did ten years ago, and do it with a smaller pixel size, but the demand for that is just not there in today's market. A manufacturer must want it.

More pixels, more DR, better ISO, video. That's what the larger market wants.

I don't think it's likely that Rodenstock will make a new wide angle lens line to fully support Sony's sensors. This is what you get, take it or leave it. More likely is that we get a medium format mirrorless from Hassy and/or Phase One, like a big A7r with a flange distance not as short as tech cams but shorter than the MF-DSLRs, and new wide angle lenses for that will be the future high-end king. By then resolution will be so high that movements will be done in post-processing, keep camera level and just crop, or tilt the camera and keystone correct. Lens tilt? Focus stack instead, automatic in the body like already on the XF.

I think classic tech cam and "large format style" photography could be commercially successful into the future though, but the manufacturers would need some Leica M thinking. Market it as a simple classic, support it with proper digital hardware. Tech cams has got stuck in market for the rational professional, where all is about technical performance and efficiency in workflow. When it's not as efficient or excel as much technically any longer, it fades away.