Phase One IQ3 100mp. Shipping now.

[torger]

Yes, it's indeed possible, and then we can use an Xf and regular lenses and a stitching head and be done with it.

My heart beats for the large format single shot shooting style though, and I would hate see it fade away.

[Theodoros]

Hi,

I am not a Hasselblad Multishot expert, but my understanding is that for 200 MP you do 16 exposures while shifting the sensor a half pixel. Now shifting the sensor a half pixel between exposures gives you for point samplings per pixel. So, what happen is that you split the pixel in four subpixels, each half the size of the original pixel. You don't loose full well capacity as you shoot four different exposures.

But, what happens is that you effectively reduced pixel size. You actually gain a lot, because you also take 16 exposures and can eliminate much of the Bayer interpolation, definitively a good thing.

Yes, you double Nyquist limit, but it is by virtue of emulating half the pixel size. So, to deliver 200 MP, your lens actually need to resolve those 200 megapixels, it is as simple as that, and if they do that, they will work just fine with an 100 MP back.

Best regards
Erik

I thought that's what you (and others) presume it happens, but it's not... You don't "split a pixel in four"... what happens instead, is that you create a new (real - true color) pixel of 9μm that has another 8 around it... Out of these 8 pixels, the upper, lower and side four are also new pixels (the five of them forming a cross), while the corner four are identical to the ones that would exist if the same back was shot at 4x mode, (but moved when mapping the final 16x image a row or a column away from what would be their neighboring column or row if shot at 4x mode). Don't forget that each one of the 9 positions of pixels in the above example is a result of four shots each, resulting in a true color one... None of the resulting pixels of the final file uses existing (position or color) information used by a neighboring pixel...

EDIT: I had to edit this, as in the first place I've put things the other way around than there is (by mistake). In the first post I had the pixels that are identical to the ones that would exist if the back was shot in 4X mode where the ones of the 16X mode are created and vice versa.

[adammork]

So maybe a slightly different workflow can also be an alternative.

Not really if you try to make a living of architectural photography - it maybe work in theory, but as an alternative to direct shifting, I think not....

[alatreille]

I think this is already a common workflow in certain situations for many AP's - though only through using a shift lens to stitch. ie - using the 17mm TSE for an ultra wide shot.

Whilst it is a viable method, it comes with it's own set of difficulties as compositing of exposures is also common practice.  ie - time on the computer goes up! 

Yes I understand, but there are other possibilities than using optics that create issues that cannot be accommodated for. A simple stitch of two or three images can sometimes solve a lot. It allows to use the center of the image circle (is better quality), perhaps use longer focal length (is usually better image quality), and the stitching software can allow to use a projection method that preserves certain straight lines/edges, yet compress the extreme stretching to a perceptually more normal proportion. Leveling and squaring surfaces is also much easier and more accurate than possible with a limited preview screen size, and saves time on location (less variation in outside light conditions).

So maybe a slightly different workflow can also be an alternative.

Cheers,
Bart

[adammork]

I think this is already a common workflow in certain situations for many AP's - though only through using a shift lens to stitch. ie - using the 17mm TSE for an ultra wide shot.

Whilst it is a viable method, it comes with it's own set of difficulties as compositing of exposures is also common practice.  ie - time on the computer goes up!

I stitch a lot, direct stitch where you are shifting the lens or back, here you can, somehow, easily combined different frames of the images so you can place people or not for instance....but if you start stitching around the nodal point, without shifting, as suggested here, you are asking for all kind of troubles with lines not lining perfect up and so on. If you have all the time in the world you can maybe make it work, but in a commercial workflow - that would be a challenge.

[Ken R]

In architecture photographers use ultrawide not because they want to, but because they must due to limited space.

Yes, also because it is faster to use an extreme wide angle lens rather than a longer lens and stitch. Both can yield the same effective angle of view / coverage but one adds several steps in the workflow. Also in instances where there are elements moving the frame one shot capture is preferred for obvious reasons.

I am hired to shoot dozens and dozens architecture projects each year. Most involve interiors with available light. The jobs do not pay very well (quite a few people get paid even less than me in the local market and those drive the price down) and to be profitable one must be very fast and efficient and of course produce really good images consistently. Partly to be able to have a fast turnaround and also because sometimes access to projects is very time limited.

I mainly use my Canon cameras due to them producing good color in mixed and low light situations and also due to the high quality wide angle lenses available. For interior shots the 14mm L II lives on my 5D3. The new Canon DSLRs also require basically no dust spotting in post (the Nikons and Sony, and of course MFDBs still require a good bit of it). Might be a stupid thing but saves a ton of time although today's software has improved on this tremendously. The dynamic range is good enough for 99% of the time. I do wish for better in a few instances though. (Hence why the D810 and the A7R/II are so popular for Arch. photos).

I have used my tech camera and IQ160 back for some specific jobs and of course results have been stellar.

But generally a DSLR system is preferred due to the faster workflow, lens selection and the integrated body.

I would love to use a XF and a IQ150 for Architecture but one key element is missing, an ultra wide angle lens. Yes that can be solved (substituting the XF) with a ALPA FPS (or Hartblei or Arca Body and an FPS plus adapters) and a Canon 24mm TS-E II (which I own) or even the 17mm, but the cost is prohibitive and you loose the convenience and speed of working with a DSLR.

A Leica S (007) is a good alternative though. Same with an H5D. Both systems have a good 24mm lens available. The 645Z would be a great choice if it had an extreme wide angle lens available.

Of course image quality wise the 100MP IQ back is most likely above all. Would be awesome if Phase made a 24mm lens for the XF system. That would round out the system nicely. The new 35mm seems impressive.

Again all my comments are for Architecture and Interiors photography.

[torger]

It would be a pretty simple engineering challenge to design a motorized head (already exists) so you just press a button to shoot all frames, and get a preview image in the tethered computer, choose perspective and crop, and voila an auto-stitched virtual raw that you then process in your favourite raw converter like any other raw. As far as I know it doesn't exist software to make the workflow this way but if there's a demand for it this type of solution it can be realized.

If some real development effort is put into this a highly effective stitching solution could be made that wouldn't require that much computing time, with reservation for lots of moving stuff.

I'm a software engineer and I would enjoy developing such a solution, but I would not really enjoy using it for my own photography. I like the hands-on "mechanical craft" of photography and at some point the image making becomes too much of a technical exercise. It's all in my head of course, cropping and shifting on the computer screen rather than on directly on camera is not really a creative difference, but I just like the latter better.

[torger]

There's also the Hasselblad system with the tilt-shift adapter which gives you quite many tilt-shift focal lengths. That you get movements recorded in the exif data and automatic lens corrections is a neat and unique feature. As it's a tele converter solution the quality is not Rodenstock but the automatic lens corrections make up some, and there's no LCC need.

That could be an interesting alternative for me personally as I like to have many focal lengths with tilt-shift capability, and I happen to have a H4D. The system would be super-expensive and super-heavy compared to my current tech cam system though.

I'm not really into maximizing resolving power at all costs, the shooting experience and compositional flexibility must be there too.

[alatreille]

I stitch a lot, direct stitch where you are shifting the lens or back, here you can, somehow, easily combined different frames of the images so you can place people or not for instance....but if you start stitching around the nodal point, without shifting, as suggested here, you are asking for all kind of troubles with lines not lining perfect up and so on. If you have all the time in the world you can maybe make it work, but in a commercial workflow - that would be a challenge.

It's all about lines... ;-)

[AreBee]

torger,

...a highly effective stitching solution could be made...

360Precision - Absolute

[BernardLanguillier]

It's all about lines... ;-)

I stitch a lot, often buildings. There is zero issues stitching perfectly lines to the pixel when rotating a good lens around it's entrance pupil when using good pano software such as PTgui. This is simply a non issue. A recent example.



I also own a pro motorized head (Clauss Rodeon) that I can control with my iPhone or my soon to arrive Surface Pro 4. This head is super robust and can support heavy body/lens combos easily.

But there is still a potential problem with stitching in some cases when some vertical elements (typically a supporting column) are located fairly close to the camera. As the camera rotates around an horizontal axis when switching between rows, the distance to the vertical column changes, which results in differences of sharpness btwn the center and top/bottom of the column. Stopping down helps a lot and focus stacking fix this issue but can create some alignement issues. Here is an example where this issue occurs.



I am sure that those who tried stitching seriously know about this problem, but this and the additional time it can take are IMHO the main issues with the use of stitching for architecture work. Time can be discussed as it may in fact not be that much more time consuming with an optimized workflow.

But this isn't the point I believe. I was hoping to be able to shoot with a FF DB on a tech camera using live view without having to care about weird technological issues with color impacts. In short I was dreaming of a real alternative to a 4x5 camera. From these initial reports it would seem that the IQ3 100MP still struggles to make this a reality and, if confirmed, this would be disapointing considering the price of such a system.

For what I do it pretty much kills the value compared to a potential Pentax 645X using the same chip at one quarter the price. Strobe shooters will still like the central shutter lenses of course, so this will probably not make much of a difference for a majority of IQ3 100MP shooters.

Cheers,
Bernard

[AreBee]

Bernard,

There is zero issues stitching perfectly lines to the pixel when rotating a good lens around it's entrance pupil when using good pano software such as PTgui. This is simply a non issue.

Exactly.

[Theodoros]

IMO, the best camera to shoot architectural interiors is by far the Roundshot 220VR... One shoots 120mm (or 220 if he needs ultra panoramic look done with telephoto) film with it, but with a Nikkor 17-35mm f2.8 used at only two or three focal lengths (say 21, 27 & 35 - nodal point pre-found and noted for ever), one can shoot whatever wide angle he wants, at whatever height he wants and as for detail... well... you'll never (and I mean never) miss some... For exteriors one may need the Zeiss 15mm f2.8 additionally and the shift adjustment added (usually a standard), but then the result will be very hard to beat and there will be no scene that wouldn't prove "a piece of cake" for the system... Killer camera! It starts where hi-end photography ends....

[ben730]

IMO, the best camera to shoot architectural interiors is by far the Roundshot 220VR... One shoots 120mm (or 220 if he needs ultra panoramic look done with telephoto) film with it, but with a Nikkor 17-35mm f2.8 used at only two or three focal lengths (say 21, 27 & 35 - nodal point pre-found and noted for ever), one can shoot whatever wide angle he wants, at whatever height he wants and as for detail... well... you'll never (and I mean never) miss some... For exteriors one may need the Zeiss 15mm f2.8 additionally and the shift adjustment added (usually a standard), but then the result will be very hard to beat and there will be no scene that wouldn't prove "a piece of cake" for the system... Killer camera! It starts where hi-end photography ends....

Sorry, I completely disagree, but that's out of topic.

[eronald]

What a useless discussion - this back will sell to people who know it will work for them, and amateurs. A pro with CA or eveness issues is not going to order one of these 50K units in the hope she might find a lens that works when shifted. Besides which anyone smart enuff to ask the question is already stitching the Sony or doing multishot with a smaller back.

Edmund

[BernardLanguillier]

What a useless discussion - this back will sell to people who know it will work for them, and amateurs. A pro with CA or eveness issues is not going to order one of these 50K units in the hope she might find a lens that works when shifted.

What a useful comment!

Cheers,
Bernard

[eronald]

What a useful comment!

Cheers,
Bernard

I feel so honored talking to people who buy $50K gear 

Edmund

[BernardLanguillier]

I feel so honored talking to people who buy $50K gear 

As of now, people who are considering buying 50K$ gear. Which is the whole point of this discussion.

You apparently think that the only people buying such equipement are super rich and therefore don't really care about the value delivered since the cost is insignificant, I have no problem with that being you opinion.

But I would just appreciate if you could avoid contributing turning this into a self realizing prophecy by not giving this camera the chance to prove its ability to perform up to its price tag. Which you are trying to do by killing a relevant technical discussion.

Cheers,
Bernard

[torger]

Due to a current Lcc bug in C1 it's easy to get confusing results currently. It's not yet clear where the shift limits are, but it looks quite good for 15mm on 40HR which will bring you far.

That the Rodenstock Digarons lenses and sensor don't match is clear and that crosstalk therefore is a factor and you don't get the full 90mm IC on the wides is 100% clear. But many shooters can do with a fair bit less, some say they need only 8mm of shift and for those this can work out just fine.

It looks like it has slightly less issues than the IQ250 but more than 80Mp Dalsa.

My reports can come out as negative as my personal interest is in larger shift ranges, and I am more concerned than most about crosstalk related color fidelity reduction

I am also concerned by the raw fact that the highest end sensor actually doesn't match Rodenstock designs and we need to push the envelope into the unknown compromise space like this. It's not how it should be and I am not particularly fond of that we as photographers need to do these evaluations. If it's 90mm IC it should work predictably to the edge and that was Rodenstock's intention. They couldn't predict Sony's sensor design though.

[torger]

I'm also not happy about Phase Ones solid silence about these issues and that they never mention crosstalk -- crosstalk is an uncontroversial standard term in image sensor design and it's obvious that the special issues we have here is related to that. It's easily explained too. I don’t understand why Doug and other representatives are so reluctant to using this term. It's almost like they want photographers not know so they don't know what to look for and therefore buy without knowing what compromises they make when shifting.