Phase One IQ250 Tech Cam Testing (vs IQ260 vs IQ280) by Digital Transitions

[Bart_van_der_Wolf]

Also I'm really surprised that the IQ250 appears sharper - some of the text on the sign is clearly readable but not so much on the IQ260.

Hi Richard,

The denser sampling pitch of the  IQ250's sensor array will extract higher resolution out of the same lens. There may have been focus related issues that additionally contributed, but I assume that Doug tried to keep focus as constant as possible.

Cheers,
Bart

[torger]

Here is a copy converted in RT 4.0.11.79.

Note that RT does not have any color matrix for the IQ250 yet, which means that it just uses a default one, I think it assumes (=pretends) the RGB primaries of the camera are the same as sRGB. When we have a proper matrix the default color will be better. The IIQ format does have built-in matrix, but I'm not sure how that should be interpreted, it provides much too saturated color if just used straight away. (RT does not use the built-in matrixes for the IIQ format currently.)

[Doug Peterson]

Thanks very much for this comprehensive first test, Doug. Having just bought an IQ260, I'm slightly depressed to see just how well the IQ250 is controlling noise (and the 260 not, especially shifted). Also I'm really surprised that the IQ250 appears sharper - some of the text on the sign is clearly readable but not so much on the IQ260. Will await the fuller colour tests, but on this evidence CMOS full frame can't come quick enough for me.

Do note that the noise reduction effects the 250 and 260 differently. Feel free to download and play with the raws with different levels of noise reduction.

Also regarding detail rendered keep in mind the IQ250 is doing 50mp in a 1.3 crop, which means each pixel is slightly more "zoomed" in on the subject than the IQ260 which does 60mp in a 1.0 crop. What this means is if you use the same lens the IQ260 will see significantly more of the scene but have just a hair less detail on a particular subject as viewed at 100%. This advantage of the 250 would of course be entirely negated if you chose a slightly wider lens for the 250 to account for the smaller sensor.

[Richard Osbourne]

Thanks Doug, good advice. I'll have a play with the raws. Still getting used to the IQ260 after a couple of years with the P45+. They don't seem very different so far, apart from the user interface of course. I bought the Schneider 28XL and have mostly had better results WITHOUT the centre filter, which causes a lot of ghosting and flare and some detail loss. Perhaps the centre filters won't change the results of this test much either.

[Doug Peterson]

Thanks Doug, good advice. I'll have a play with the raws. Still getting used to the IQ260 after a couple of years with the P45+. They don't seem very different so far, apart from the user interface of course. I bought the Schneider 28XL and have mostly had better results WITHOUT the centre filter, which causes a lot of ghosting and flare and some detail loss. Perhaps the centre filters won't change the results of this test much either.

Most of the critiquing of the noise in the 260/280 files has been about the noise in the heavily shifted areas.

With both these backs a multi-second exposure produces excellent results if the file is used as-is, but in both cases pulling up the shadows a few stops (as is the case with the LCC correction of the falloff with several of these lenses) will strain a multi-second file. As you can see with the IQ250 it seems a multi-second file is just as pliable as a short exposure.

Notably the IQ260 can use it's long-exposure mode wherein the amount of noise is increased modestly from native ISO50, but remains more-or-less constant from a short exposure out to a very long exposure. So this shutter speed range (5-10 seconds) is really the one that strains the IQ260 the most - longer than you get IDEAL results out of (you still get very very good results - check out the unshifted image, or request from me one of the shots where I shifted the IQ260 and then changed exposure to adapt for the falloff, lowering the load on the LCC routine) but too short to justify using the Lone Exposure mode.

I suspect if we repeat this test in a sub-second landscape or even a less dim interior that the 260 would hold it's own significantly better.

I also suspect if we had used Center Filters on the 32, 40, 47, and 60 - or changed the exposure for each lens-back position to optimize exposure for that are of the image circle (rather than holding the shutter speed constant) that the results from all the backs would have improved, especially the 260/280. As you say though Center filters come at the cost of having to worry more about flare. The Arca and Cambo compendiums help a lot with that, but there is, regardless an increased hassle factor.

But if nothing else this test shows the 250 can take an absolute beating (multisecond exposure of a mostly-back-lit brown-red wood in the corner of an image circle with heavy fall off) and come out with really strong results. The same can be said of the 260 and 280 in shorter exposures.

If it weren't accumulating 10-15 inches of snow outside (and I weren't' so much a wus) I'd probably be doing more testing outside to provide a broader perspective.

Good news is we just completed the capture part of a skintone test of IQ250 vs IQ260 at each back's complete ISO range. We'll be working with Douglas Sonders to post that test along with the raws so everyone can play to their hearts delight. Probably won't be processed/evaluated/uploaded/posted for a week or so. It'd probably be faster but people keep coming in wanting demos of the IQ250, darn them! (I'm kidding of course, I enjoy working with photographers in person much more than processing images in a dark office).

[Richard Osbourne]

This kind of interior shooting is the kind of thing I do a lot of - long exposure, no flash and pulling up the shadows in post to avoid bracketing exposures. As you say, Doug, the beating the IQ250 files are taking in the shadows is impressive. Which back has the most accurate colours? They are quite different from one another on this showing.
Not sure comparing two different shots of snow will tell us much about colour unfortunately. 

[torger]

I've researched sensor tech a bit and tested the files with my own LCC algorithms, and I think the IQ250 suffers quite a lot from pixel crosstalk on shifted tech wides. Pixel crosstalk means that light comes in through one color filter but angle is so low so it crosses over and gets registered in a neighbor photodiode. Red could be registered in green etc.

The most noticeable artifact of pixel crosstalk is actually not blurring of detail (the blurring of a shifted lens is larger, even great ones like the 32HR) but instead desaturation of colors.

This means, the more you shift your wide, the more desaturated colors will be. This is quite noticeable in the 32HR stitched jpeg (colors get kind of brownish and dull). Saturation can of course be restored to some extent with manual effort in photoshop (LCC can't correct it!), but this puts further stress on the dynamic range and color separation will suffer.

In other words, if color fidelity is excellent in the center of your wide angle lens, you may get color issues on the sides of your stitched ultrawide shot. This is probably something interior photographers need to consider.

As desaturation happens gradually it's hard to say anything definitive about the wide angle usability. I'm considerably less enthusiastic about combining the IQ250 with tech wides now after some analysis than I was when I first saw the test images.

[Fine_Art]

In other words the lens has to be designed to deliver highly collimated light so that the added angles of shift/tilt do not smear the light across the sensor surface at these tiny distances. Maybe enhanced microlenses can help.

[torger]

In other words the lens has to be designed to deliver highly collimated light so that the added angles of shift/tilt do not smear the light across the sensor surface at these tiny distances. Maybe enhanced microlenses can help.

More retrofocus in the lenses (Canon's TS-E 17 and 24 works great), putting photodiodes closer to the surface (backside illumination process) and possibly enhanced microlenses and lightpipes. And of course making the pixels larger would help too.

Maybe it's possible to hinder crosstalk by putting in some "walls" between pixels so light can't cross under the filter. From center to close to the edge of the image circle of the 32HR the IQ250 loses ~4 stops of signal due to pixel vignetting, I don't know how much would further be lost if crosstalk was eliminated by just letting the light smash into walls, but I think it's a fraction of a stop, otherwise desaturation issue would be larger than it is.

I don't think it's a great idea to solve the problem only with optics, ie make tech cam lenses as retrofocus as DSLR lenses, as it will then be very difficult to keep the performance edge in terms of sharpness and lack of distortion tech cams are known for, and lenses would be even larger heavier and more costly than the current 32HR.

[ErikKaffehr]

Hi,

Don't forget that there is a natural loss of illumination according to cos^4 that accounts for two stops alone on a 32 mm.

But you are probably right that LF-lenses and the Sony sensor are not created for each other. Leica M has a lot of issues with wide M-lenses and wide M lenses don't work well on the A7r.

Best regards
Erik

More retrofocus in the lenses (Canon's TS-E 17 and 24 works great), putting photodiodes closer to the surface (backside illumination process) and possibly enhanced microlenses and lightpipes. And of course making the pixels larger would help too.

Maybe it's possible to hinder crosstalk by putting in some "walls" between pixels so light can't cross under the filter. From center to close to the edge of the image circle of the 32HR the IQ250 loses ~4 stops of signal due to pixel vignetting, I don't know how much would further be lost if crosstalk was eliminated by just letting the light smash into walls, but I think it's a fraction of a stop, otherwise desaturation issue would be larger than it is.

I don't think it's a great idea to solve the problem only with optics, ie make tech cam lenses as retrofocus as DSLR lenses, as it will then be very difficult to keep the performance edge in terms of sharpness and lack of distortion tech cams are known for, and lenses would be even larger heavier and more costly than the current 32HR.

[torger]

Don't forget that there is a natural loss of illumination according to cos^4 that accounts for two stops alone on a 32 mm.

Yes, and the test did not use a center filter (which would improve that), but losing 3-4 stops of the signal seems to be the least of the problems as this sensor lots of dynamic range to pick from.

The different angular response on the greens and the crosstalk seem be show stoppers though, at least for the 32mm. The 40 could work but I'd want to do more testing.

If color fidelity only can be had in the central part of the image circle, the meaning is lost. I certainly prefer color fidelity over resolution, to enjoy the sharpness of a wide angle tech cam lens you need to get up close, but color you see on a distance.

Sharpness is easy to evaluate, just pixel peep. Color is harder to evaluate though, and the gradual falloff in performance is kind of nasty, it may look good on one subject but worse on another. Therefore I would be very careful before investing in a IQ250 to be used on a tech cam.

[ErikKaffehr]

Hi,

No argument, I essentially agree.

I would also agree that current designs are not very friendly to centric WA designs. This is quite obvious with Leica, who do a lot of SW corrections on both the M9 and the. Leica even developed (paid for development?) for their own CMOS sensor. The CMOSIS sensor has shallower wells than the Sony sensors.

For me the IQ-250 is not very interesting, to begin with, I cannot afford it. Also my route to MFD is using old Hasselblad lenses, with a Hartblei HCam on the distant horizon, so crop factor matters. MF folks always used to say that CCD-s were much better than CMOS designs, better DR, better colour and the great advantage of sixteen bit data, not that I believe any of that. Statements like that really undermine MF credibility for me. So it is interesting to see how the IQ-250 performs.

I would say that LiveView is probably nice for technical cameras, especially if combined with good peaking. LiveView is one of the things I miss on the P45+. I know that it has LV, sort of, but I don't shoot in studio (except tests) so I have little use for it as I cannot have live view on the back.

If I stay with MF I guess I would get a Hartblei HCam. Use my V-series lenses with a Mirex TS adapter and get both CanonTS lenses. But I may just skip MF and stay with Sony.

I am somewhat ambivalent about MF right now. "Where is the beef?" is the question I ask.  

Best regards
Erik

Color is harder to evaluate though, and the gradual falloff in performance is kind of nasty, it may look good on one subject but worse on another. Therefore I would be very careful before investing in a IQ250 to be used on a tech cam.

[Fine_Art]

Note that RT does not have any color matrix for the IQ250 yet, which means that it just uses a default one, I think it assumes (=pretends) the RGB primaries of the camera are the same as sRGB. When we have a proper matrix the default color will be better. The IIQ format does have built-in matrix, but I'm not sure how that should be interpreted, it provides much too saturated color if just used straight away. (RT does not use the built-in matrixes for the IIQ format currently.)

That's ok, I did my own correction of the "primaries" based on what I thought was natural and flattering color. Natural comes first. I have not been in that building so if my less saturated interpretation is off I retract my use of the word enthusiastic. What do other people see? Nobody else is throwing in an interpretation. This is a rare opportunity, you have a very nice subject, a very high end camera system, at least 2 raw converters that will take it on. C1 and RT. People should be looking at what they can achieve vs their regular system.

For me, I am like Bernard, I would stick with stitching. D600 + Nikon 85 1.8G would handle this subject well. So would Sony A55 with 50 macro. Color accuracy would be a bit less, DR would be noticeably less, requiring HDR. Still, the job would get done.

The big test, as Doug has already stated, is when you put this system in a situation where you have to do it in one shot. Fast changing light or moving subject. Give us a green flash over mountains at dawn to play with!

[ErikKaffehr]

Hi,

I would guess that the D600 has similar DR to the IQ-250. Similar technology and larger pixels. Stitching also gives more pixels.

Sharpness from the IQ-250 on the technical camera is impressive, but comes with extensive aliasing.

Best regards
Erik

That's ok, I did my own correction of the "primaries" based on what I thought was natural and flattering color. Natural comes first. I have not been in that building so if my less saturated interpretation is off I retract my use of the word enthusiastic. What do other people see? Nobody else is throwing in an interpretation. This is a rare opportunity, you have a very nice subject, a very high end camera system, at least 2 raw converters that will take it on. C1 and RT. People should be looking at what they can achieve vs their regular system.

For me, I am like Bernard, I would stick with stitching. D600 + Nikon 85 1.8G would handle this subject well. So would Sony A55 with 50 macro. Color accuracy would be a bit less, DR would be noticeably less, requiring HDR. Still, the job would get done.

The big test, as Doug has already stated, is when you put this system in a situation where you have to do it in one shot. Fast changing light or moving subject. Give us a green flash over mountains at dawn to play with!

[torger]

I am somewhat ambivalent about MF right now. "Where is the beef?" is the question I ask.  

If one like you and me not use the latest so we don't have the single shot resolution advantage to enjoy I think the joy must come from the camera. The back as I see it is just about getting adequate performance to a not-too-high price. And then it may not be using a better camera, a Hassy V or Mamiya RZ is certainly not easy to operate compared to a consumer DSLR, but it's different, a bit nostalgic and quite fun. I like my Linhof, and while I do enjoy to study new technology I don't really own much of it, I like being able to have this camera and not feeling stressed by the upgrade race, and I like the slower workflow.

However, using a Hassy V and always missing focus can be frustrating, and focusing on ground glass on my Linhof is not for all. So while I may enjoy my camera others could hate it, we're all different. If one does not really enjoy using the camera the back is attached to, ie would rather use the convenience of a DSLR, I think selling the back is a wise decision. It can't be kept just based on image quality.

[Fine_Art]

Hi,

I would guess that the D600 has similar DR to the IQ-250. Similar technology and larger pixels. Stitching also gives more pixels.

Sharpness from the IQ-250 on the technical camera is impressive, but comes with extensive aliasing.

Best regards
Erik

I think so too. The bit less DR and Color accuracy comment was for the A55.

If you look at the sorted list of cameras at DxO the D600 is #3 behind the D800E and D800. The Phase IQ180 is below the D600.

[Fine_Art]

I have deleted the alternate conversion i posted. It doesn't look like any comparisons will be coming out.

The file is quite impressive, as it should be from a camera costing as much as an SUV.

[BernardLanguillier]

All in all I have the feeling that the acceptance of the DxOMark results as a credible measure of real world camera performance, in particular DR, has been increased by these tests.

Cheers,
Bernard

[ErikKaffehr]

Hi,

I don't know about DxO mark but I see that Sony Exmoor works as Sony Exmoor even in MFD, and it seems to work really well.

Best regards
Erik

All in all I have the feeling that the acceptance of the DxOMark results as a credible measure of real world camera performance, in particular DR, has been increased by these tests.

Cheers,
Bernard

[billrcahill]

Great test. I am coming from an Aptus II 12 R. I shoot product on a sinar P2. using a bass V mount. I am wanting to upgrade and get a MF system as well. With the release of the 50mp back, the higher iso is attractive. I can't seem to shoot my back above 100 without it starting to fall apart. All I really need is 400, and that would help a lot. I don't see much loss of detail between the 80 and the 50, besides the loss in megapixels. Do you see any reason not to downgrade to 50 megapixels?