IQ250 movement tests on the 23/32/40/70/90 - desaturation and mazing artifact

[voidshatter]

Rippling on Dalsa sensors is partially or entirely removed when using the LCC tool in Capture One using the optional setting "Technical Camera Lens".

Capture One v8.1 still fail to completely eliminate tiling on the Dalsa CCD or rippling on the Sony CMOS. If you apply micro-contrast heavily during post-processing for flat areas such like the sky you still observe artifact. You can even observe this on many of those offical sample images published by Phase One.

 

[tjv]

Is the adapter plate the only problem for the demo? Isn't it possible to get a V-mount adapter plate for your sliding back?

Because I use film, I don't have a sliding back. I just swap the Silvestri GG out for the Horsehan film back.

 

[torger]

Capture One v8.1 still fail to completely eliminate tiling on the Dalsa CCD or rippling on the Sony CMOS. If you apply micro-contrast heavily during post-processing for flat areas such like the sky you still observe artifact. You can even observe this on many of those offical sample images published by Phase One.

The challenge with the tiling and ripple is that it does not *exactly* match what you get on the LCC (due to crosstalk which is color-dependent), so you can't just use the LCC to cancel it out, you need to apply a filter too. I do that in Lumariver HDR, and I'm sure C1 does it too, but even then there will be a residual ripple left and the only way that seems to work to remove that is to add in noise and that neither my algorithm or C1 do.

For 99% of post-procesing it's a non-issue though. The only folks I've seen that have problems with it is those that do black and white processing with very high contrast increases, and then you can remove it with retouching.

 

[AreBee]

voidshatter,

...movement along the shorter edge of the sensor should not exceed 15mm... (actually, 12mm to be conservative for the 32HR as suggested in the OP). This would work to avoid mazing if my hypothesis is true.

And, if true, implies that mazing would not be visible with a full frame CMOS sensor, but that the range of available shift would be limited only to several millimeters.

 

[voidshatter]

voidshatter,



And, if true, implies that mazing would not be visible with a full frame CMOS sensor, but that the range of available shift would be limited only to several millimeters.

If the fullframe CMOS sensor is going to be based on the same tech as of the IQ250, and if my hypothesis of mazing based on my systematical tests is proven to be true, then yes you may not be able to utilize full image circle of the 32HR, while you could make almost full use of the image circle of the 23HR, and make completely full use of the whole image circle of the 40HR.

Note that the above is still only a rough guide and I will update should I discover anything new, or if Doug could provide a counterexample to invalidate my hypothesis.

 

[AreBee]

voidshatter,

If the fullframe CMOS sensor is going to be based on the same tech as of the IQ250...

You mentioned in a previous thread about a 120MP full frame prototype. A pixel pitch as per the IQ250 would return 77MP, full frame.

Hopefully the prototype is not based on the same tech...

 

[voidshatter]

voidshatter,



You mentioned in a previous thread about a 120MP full frame prototype. A pixel pitch as per the IQ250 would return 77MP, full frame.

Hopefully the prototype is not based on the same tech...

Current prediction:

a) D7000/D800E/IQ250 tech in 645 format: 80 MP, current Rodenstock HR usable

b) NEX7/D5300 tech in 645 format: 120 MP, current Rodenstock HR not usable

c) NX1 tech in 645 format: 141 MP, current Rodenstock HR and Schneider XL probably usable (but DR will be less)

Personally I would hope for the last one but given that it's still in APSC size as of 2015 it might still take 4 years or longer to realize (it took 4 years from D7000 to IQ250).

 

[tjv]

Thanks to the very generous support of Paula at Linhof Studio, I am able to test the CFV-50c for myself in the coming weeks on my own Linhof Techno. I'm interested to see how my Rodenstock 55mm APO-Sironar Digital performs with shifts. The only other lens I currently have is the 90mm HR-W (blue band), but that focal length doesn't interest me at all on this format. I have realistic expectations but high hopes...

 

[Paul2660]

Current prediction:

a) D7000/D800E/IQ250 tech in 645 format: 80 MP, current Rodenstock HR usable

b) NEX7/D5300 tech in 645 format: 120 MP, current Rodenstock HR not usable

c) NX1 tech in 645 format: 141 MP, current Rodenstock HR and Schneider XL probably usable (but DR will be less)

Personally I would hope for the last one but given that it's still in APSC size as of 2015 it might still take 4 years or longer to realize (it took 4 years from D7000 to IQ250).

One question I keep asking is: if you make a 80 to 100 MP CMOS a sensor based on the technology of the IQ250 this would the pixel pitch get even smaller than the current IQ250? If so then this new Full frame CMOS would have less than desirable performance with tech lens camera especially when movements are needed?

Wouldn't any full frame sensor have to have an even smaller pixel pitch?

Paul

 

[ErikKaffehr]

Hi,

The IQ 250 upscaled to "full frame" 645 would be around 85 MP, keeping the pixel pitch. I don't know which vendors Phase One is working with, but would be feasible for Dalsa to make a 100 MP sensor for Phase.

Personally, I like small pixels, but I also realise that small pixels don't work well with large beam angles, like the ones coming from symmetrical lenses of short focal length.

It may be possible that shallower wells work better, but I guess that shallower wells may give up a bit of DR.

I would guess that one problem with MF is that sensor design is driven by developments where the optical system and sensor are optimised as a whole, like mobile phones.

It may be possible that a vendor like Phase One would be able to designs a sensor with say Dalsa that takes technical cameras into account. But it is a bit complex. Phase One is also a camera maker, and technical cameras is not a dominant piece of the market for Phase.

I guess we have to wait and see to find out.

Personally, I am pretty sure that small pixels are optimal for the DSLR/Mirrorless market. With technical cameras it is a bit different, due to he large and variable beam angles.

I would guess that we may see more complex sensor designs with shallower pixels combined with increased full well capacity, that would solve many of the issues we face today.

Best regards
Erik

One question I keep asking is: if you make a 80 to 100 MP CMOS a sensor based on the technology of the IQ250 this would the pixel pitch get even smaller than the current IQ250? If so then this new Full frame CMOS would have less than desirable performance with tech lens camera especially when movements are needed?

Wouldn't any full frame sensor have to have an even smaller pixel pitch?

Paul

 

[voidshatter]

Mazing occurs because the demosaicer expects green 1 and green 2 to be the same on a uniform surface, but due to crosstalk they separate and then the demosaicer starts to make up detail. The greens differs because one of the greens get most it leaks from the neighboring red and the other from the blue. Of course if you're lucky and the color you're shooting has equal red and blue content (on the raw level), green1&2 will not separate.

Funny enough... I didn't really get into pixel peeping until I moved into the world of MFDB. Today I went back to my old files taken by a Nikon D4 and a Nikon 16-35mm f4 VR (link to download RAW as evidence), and guess what - I observed mazing! This isn't really a high resolution sensor or a tilt-shift lens. I guess mazing is just a fact of life, but we are too harsh about the MFDB!

 

[tjv]

Strange, I've never seen that in any Nikon files I've shot. How hard did you push the file?

 

[torger]

Funny enough... I didn't really get into pixel peeping until I moved into the world of MFDB. Today I went back to my old files taken by a Nikon D4 and a Nikon 16-35mm f4 VR (link to download RAW as evidence), and guess what - I observed mazing! This isn't really a high resolution sensor or a tilt-shift lens. I guess mazing is just a fact of life, but we are too harsh about the MFDB!

I think that's a special case with some reflection phenomenon occurring to due to the very sharp light source on the edge of a ultra-wide. There's no mazing on the opposite side. I also think the effect may be strengthened due to the narrow band light source. I would be surprised if that effect occurs in a scene with more normal light.

 

[Y Sol]

With NX2 (version 2.4.6) it looks different. Here are 2 screenshots.

 

[narikin]

So... just to turn this on its head - if I wanted the mazing artifact to appear - what is the best conditions to make it happen?!

I see some interesting examples in your initial post, and I'd like to recreate it.

 

[torger]

If you have an IQ250/CFV-50c the simplest way is to use a symmetrical wide angle and shift it a bit, and if you have retrofocus rodie wides then shift it quite a bit

Different demosaicers have different sensitivity concerning green split, if you want something really sensitive try "DCB" in RawTherapee, then mazing will appear early on. Capture One and Phocus demosaicers are quite robust, but eventually break down.

 

[thrice]

Current prediction:

c) NX1 tech in 645 format: 141 MP, current Rodenstock HR and Schneider XL probably usable (but DR will be less)

Personally I would hope for the last one but given that it's still in APSC size as of 2015 it might still take 4 years or longer to realize (it took 4 years from D7000 to IQ250).

The NX1 does not contain a Sony sensor, AFAIK Samsung have no desire to enter the medium format sensor fabrication business.

 

[kdphotography]

But what about the Fuji GX680?

:ROTFL:

:salute:

 

[Don Libby]

Oh boy...:facesmack: :watch: