You sure? You ever shoot with any of these systems? My PhaseOne system is only a few pounds heavier than the Canon system I take when shooting high end work. Sure my Canon street setup is much lighter, but when doing landscapes there isn't much difference. Most of the paraphernalia has to do with shooting not the system, and goes along with either system.

Jeff,
I would never argue that one should not try to use the best tool for the job, but I get the impression that many photographers would have an eye on a P65+ because they want one rather than need one.

There also seem to be some serious disadvantages to the MFDB system, such as a slow continuous frame rate, (less than one frame per sec with the P65+), reduced performance at higher ISOs (or reduced pixel count as an alternative) and shallower DOF at any given F stop (for same FOV), which is not always ideal for landscapes. (Not to mention poor autofocussing).

For example, in the above panorama of the Himalayas, how would you get both the ladies and the mountains sharp, using a P65+? F22 at ISO 44 in the early hours of the morning?

Wayne,
Doesn't lower base ISO for best quality, in conjunction with higher F stop numbers to get good DoF, make use of a tripod almost mandatory in most situations, with a P65+?

Well yes and no. On a real larger format camera yes, but than again you would tilt a little and everything would be in focus at an app. around f8

If you use the phase camera than you can always go to ISO 400 which is still really good.

At the true ISO 400 (actually ISO 360 which is described as ISO 800 on the P65+) the DR of the P65+ (at the pixel level) is 2.33 stops worse than that of the 5D2 at ISO 400, and 2.67 stops worse than the D3X at ISO 400 (actually ISO 337).

Allowing a one stop difference for equal DOF between 35mm and 645 format (in fact it's greater than one stop considering the expanded full frame format of the P65+), the difference in DR between the D3X at ISO 200 and the P65+ at ISO 400 (actually 360 and nominally 800) is an astounding 3 1/2 stops.

How does the P65+ fare with autobracketing of exposure? At 0.75 frames per second, I wouldn't feel too happy trying to increase that relatively poor DR at ISO 400 by merging to HDR. Too much movement to auto-align the images.

The Achilles' Heel of the P65+ (and all MFDBs) is the poor high ISO performance. It's not surprising that Phase One exaggerates their ISO ratings.

I mean, ISO 800 actually being ISO 360. That's ridiculous!

But I don't want to give the impression I am biased in any way about this issue. I recognise that there are a few situations where the P65+ will facilitate a better job than can be achieved using the best of the current DSLRs. Nevertheless, it's a high price to pay for those relatively few situations for many of us.

Ray, where do you get these "true" ISO numbers from? Is that DxO data or from some other tests you've seen?

mmm ... when shooting for the absolute highest quality and sharpest images, does it matter what camera? I think a tripod is in order with either system - that's how I roll anyway. I would have had plenty of depth of field for the image you posted since I would have left the women out.

Wonder why everyone judges what everyone else should "need" or "want" based solely on their own perspective?

Sort of reminds me of that famous Mythbusters saying

"I reject your reality and substitute my own"

The article is well written and spells it out ... we each have our own reality of Want, need and afford. No reason to try and force everyone into the same reality we exist in.

I freely admit I have a p65 system because I can easily afford one and want one. In my mind I need one because my objective whenever I'm shooting landscapes is very large prints ... 40" is a decent starting point. No, this "need" isn't related to financial well being at all, just a personal quest for quality - perhaps by some this means it really isn't a need. If so ... I reject that reality and substitute my own

But you would you have also left the whole foreground out? The grass on each side of the photoshopped ladies is tack sharp in my panorama, taken with the now obsolete 5D Mk1.




In that case, I would ask, what percentage of your landscapes do not lend themsleves to a stitching process? Having bought a P65+ system, I would expect that you would simply use it whenever possible and disregard the option that you might have had to get a similar, or even better result using a 5D2 and stitch for the print size required. Stitching is more work, and a single shot of equal quality to the stitched image may be preferred in terms of time saved in front of the computer. But at what expense!

There are many things that I could afford but simply don't buy because I can't justify the price in relation to my actual needs, and the estimated frequency of use of the equipment. I don't need a Ferrari to get around the place and do my shopping etc., and I'm not into car racing. But if I were into car racing, that's another matter.

Whatever you want ray. I have a P65 and 5dMk2 and I know that I will always use the P65 if possible. I Only use the 5DMkII if it is very rainy or the light levels are so low that I would need a tripod but can't use one. (Out of what ever reasons.) If I had to use a 5DMkII for everything I would need to stitch 90% of every image, well you can do that and enjoy it I won't. I bought is because I enjoy working with it. I don't like to work with a small SLR and stitch up multiple images. I prefer working with a large format camera to a SLR System, I will probably sell my Canon System if there is a Leica M9.

I mean I am not saying these review sites are all wrong, but I mean take dpreview for example if you would judge the Sony a900 there you would think it sucks compared to a 1DsMk3 or d3x. I mean how can such a site publish crops which clearly show motion blur in the a900 crop samples ? I don't know how Dx0 tests all their data and I don't care. However I think you don't get the ISO results. Look at the H3D-50, Dx0 would say all ISO values from 50 to 400 are all only 50. That is true, that does not say how they act in real world. If what you think would be the case you couldn't even use a H3D in normal overcast situations.

But if I shoot a scene like yours using stitching I have to use a longer lens to capture smaller segments to match the same FoV and pixel resolution in the final file. I think the final "depth of field" between the two methods would be very similar - I vaguely recollect an article I read once which actually demonstrated they would be the same(although my 50+year old brain doesn't recollect things as well so it may have been some other point they were making). In fact, stitching is often touted as a way to decrease depth of field. I found focus stacking is really pretty simple and effective.



Don't know. Heck, I stitch p45 files, and I"m sure I will take occasion to stitch p65 files. But plenty of what I shoot would be challenging to stitch, and taking the time to use that workflow (extra time shooting, so maybe I miss another shot from a different angle), and as you mentioned all the extra time sitting at the computer doing the stitching when I could be doing something more interesting like playing golf or shooting ...

Hey ... I've been stitching since the 1Ds ... been there, done that, and it's a lot funner to not deal with it.

Nothing wrong with stitching 5dmk2 files. Go ahead, knock yourself out. You have your level of want, need and afford, I have mine. They aren't the same, never will be and neither are right or wrong. They just are. Don't make me live in your box ... I like mine.

Quite right. I wondered if someone would pick that up. I see you're alert today .

... so if you want or need three bear-breasted young ladies on a mountain top at dawn, and you can only get/find/afford one (or you only have one wife) stitching is clearly superior, even if you lug an H3D11-60 up the mountain!

If you thought hard, you could think of circumstances in which a steam traction engine would be preferable to a Ferrari.

I appreciate the benefits of pano, and I am thinking of re-taking a picture of a cornish harbour using cylinder pano stitching to get more of the "ampitheather" scene in.

Mark,
From DXO data. Just place the cursor over any of the colored blobs and you get a read-out of the measured ISO plus the manufacturer's claimed ISO, as follows:

Click to view attachment

It seems that most manufacturers exaggerate their ISO ratings, but not to this degree, surely.

I take most of my photos during trips to exotic locations. One of my main regrets is not anticipating that stitching programs like the latest version of Autopano Pro would eventually be developed and make the whole stitching process not only easier but often faultless even with hand-held shots.

I struggled with those shots of the Himalayan Dawn. I carried a lightweight, travelling tripod with ball-head which is not ideal for taking shots for stitching purposes because a ball-head can move in all directions as one tries to swivel it around. I also attempted to bracket all exposures with MLU and remote cord. That would now be a breeze with the 5D2 and LiveView, but with the 5D1 the mirror has to be flipped, then pause, before each shot. Too much movement in the scene has taken place for a good merge to HDR.

Poon Hill is one of those places I'd like to revisit.

Ray, yes thanks for the reminder of how DxO did this. I haven't been on their site for quite a while.

As for the panning technique, if you get a Really Right Stuff ballhead, you can swivel horizontally without any other movement. They also have other materials and instructions on their website for assuring the best possible fit when preparing a series of pan images.

Ray once again, they don't test what ISO they are they test what the real ISO is. I can only urge you to look at the H3D-50 again. You would see that all ISO are ISO 50 even 400. That still does noc change that ISO 400 on a H3D acts like ISO 400, even though it is not real. Same goes for phase.

so you make an incorrect statement in defense of your personal viewpoint on stitching, and then you try and pass it off as though you knew it was incorrect when you made it, but you did it as some type of test?

gimme a break ...

Wow, I didn't realize I would start a discussion that will not die. Maybe, it's time to fit a Canon scanner to the back of my 8x10.

Get a BetterLight scanning back. Less hassle (not as cheap though).

Mark,

A lot of ball heads have a panning axis, but for them to work properly one has first to adjust the legs so as to make the base of the ball head parallel to the ground. This can be a hassle. RSS does make the PCL-1 which one attachs to the ball head and then uses the ball head to level the PCL-1. The panning is done with the PCL-1. Does the RSS ball head to which you are referring have a feature that obviates the PCL-1?

I'm afraid you are not as alert today, Wayne. The issue is not specifically related to stitching but to pixel density. It so happens that the pixel density of the D3X and 5D2 is very similar to the pixel density of the P65+, so any attempt to get the same pixel count in the same FOV of scene from the same position will involve the use of the same focal length of lens and the same DOF at the same F stop, approximately.

If I were to attempt to get such a stitched image using the D3 or 5D, I would have to use a slightly longer focal length than I would use with the P65+ for a single shot of the same FOV, and consequently the DOF of the final stitched image would be slightly less at the same f stop.

On the other hand, if I were to use a 12mp Olympus 4/3rds system for stitching, which might be a better choice than a D3X when hiking up a steep hill, the DOF of the final stitched image, of same pixel count as the P65 single shot, would have greater DOF at the same f stop.

If I were to make the stitch from a Canon G10, then the DOF would be very much greater at the same f stop, but no doubt at the sacrifice of DR, SNR etc.

All these factors are related.

The obvious advantage of the 5D2 and D3X in this situation is their better performance above base ISO, in all departments. For example, If I need to use ISO 200 with the P65 to get a shutter speed fast enough to freeze the slight motion of the foliage, then in the same circumstances I could stop down more than one stop with the 5D2 for greater DOF, yet still use the same shutter speed, assuming the DXO data are reliable.

How come? The P65 ISO 200 is actually ISO 89 and the DR 10.55 EV. The 5D2 ISO 400 is actually ISO 285 and the DR slightly geater at 10.92 EV. All the other parameters addressed in the DXO tests, tonal range, color sensitivity, SNR, are also either as good or slightly better for the 5D2 at ISO 285, compared with the P65 at ISO 89.

However, when all is said and done, the final result will also depend on lens performance at the apertures chosen for each system. As a general rule, lenses with a smaller image circle designed for the smaller format camera tend to be sharper, but not necessarily at small apertures where diffraction takes its toll.

Diffraction seems to be the great equalizer.

All digital cameras have only one ISO, described as the base ISO. What differs is the way the camera handles the underexposure. I get the impression it makes little difference with DBs whether one underexposes 3 stops at, say, ISO 50 or uses the same shutter speed at ISO 400.

This is not the case with Canon and Nikon DSLRs where a correct exposure at ISO 800 will be significantly better than a 3 stop underexposure at ISO 100.

No - that's why I referred to "other materials". By the time you buy the head and the pano package you're into about USD 700+, so depending on how much of this stuff one does, either put up with the hassle of leveling the tripod, or get a PCL-1. The package also includes a plate for aligning the camera according to the optical center of the lens to minimize parallax problems.

Mark,
After that experience, I got myself a new travelling tripod, the carbon fibre, 4-section-leg, Manfrotto 190CXPRO4 with built-in level and pan & tilt Manfrotto 460MG head. It's a bit on the heavy side at 1.8Kg but at least I don't have to crouch down when peering through the viewfinder, as I had to with my other ball-head aluminium tripod which was only 5 ft high.

One series of tests I've yet to do is compare stitches of hand-held shots with properly levelled shots on a tripod, of the same scene using Autopano Pro.

Ray, re the part where "you get the impression" - not clear to me why the sensor physics should differ that much between a high-end DSLR and an MFDB. Could you explain?

And, are we talking ISO measurements per Japanese standards or per North American standards? They are different, which is why many photographers would would meter Kodak films at half their ASA/ISO and develop for more time than recommended à la Ansel Adams. Fujifilm and Ilford films always worked right at their stated ASA/ISO ratings.

The big divide between digital sensors is that some are CCDs and others CMOS. Individual photosite size makes a difference, as does color bit depth.

........and the differences are more and more complex than all of that. You can read more about the P65 sensor design on this website and in the Phase literature. One reaches a certain limit to discussions whereby actual results which test the limits of each kind of system, complemented by DxO type laboratory measurements, contribute more understanding of the value-added side of the equation than possible from cursory knowledge of sensor differences.

I'm not even sure what your point is anymore. This was a discussion about using stitching instead of buying a higher resolution camera. Your first paragraph doesn't make any sense to me. The only way I can take a scene with a dSLR that contains the same FoV and results in a file with the same pixel count as my p65 is to stitch. How is the pixel density of the sensor even relevant?. It's only affect is how many captures it will take to stitch the final file so I can match the pixel resolution of the p65 file.

To accomplish this I will have to use a longer lens in relation to the sensor size to capture the same information at the same resolution, then stitch the resulting files together. The end result is my depth of field will be quite similar from either approach, and in fact it could be the stitched fie would have less DoF, which you leveled as a criticism of the p65 (less depth of field). If you try to do this from the same location I'm not sure you can get the exact same scene including foreground and background, but then again I've never tried it, and not sure why I ever would.

$15,000 for a slow scan back with less res than a H3D11-60 (or P+65)... are you kidding?

9,000 * 12,000 pixels would give a nice 24 * 35 " print @ 360 ppi, an be useful for some landscapes, and save stitching 2 shots.

Except that those are true RGB pixel and not the result of some funky Bayer interpolation.

Cheers,
Bernard

... so, using real pixels, if there is no movement (waves, trees, people, clouds) in the shot, would you get as good a result @240 original camera pixels per print inch as you would using a flash-compatible camera (bayer interpolated) at 360 ppi?

I have been thinking about getting the 160 Mpx Seitz 617 rapid scan back for the Sinar (when they ship it) or the Red 617.

When stitching using a smaller sensor to emulate the result from a larger sensor, the pixel density of the smaller senor will determine the focal length of lens needed. The greater the pixel density, the shorter the lens needed and the greater the DoF of the stitched result, at a given F stop.

For example, if we ever get a 5D3 or 5D4 with a 60mp sensor, then you wouldn't have to stitch to get the same FOV image and same pixel count as the P65+ would provide from the same position. However, if you were to use a 50mm lens with the 5D4, you would need a 75mm lens with the P65+.

If you were to use F5.6 with the 50mm lens on the 5DMkIV, you would need to use F8 or F9 with the 75mm lens on the P65+ to get the same DoF (allowing for minor discrepancies due to the different aspect ratios of the cameras being compared, and allowing for any DoF peculiarities due to lens design and/or exceptionally close distance to subject).

If you imagine dividing the 60mp 35mm sensor into 4 parts, you would get something close to a 15mp Olympus 4/3rds sensor. To get your P65 equivalent by stitching images with a 15mp Oly, you would use the same focal length of 50mm because the pixel density is the same as the 5D4, and the DoF of the stitched result would be greater than that of the P65+ at the same F stop. Clear?

Mark,
I think we need someone like Emil Martinec to explain that, but I imagine it has something to do with the differences between CCD and CMOS design. I understand in Canon cameras, at ISOs higher than base, the analogue signal from the sensor is amplified at an early stage to reduce the effects of noise later in the processing chain.

If we refer to the DXOMark comparison of the P65+ and Canon 5D2, we can see that for each doubling of ISO for the P65+, there's approximately one stop loss of DR. At base ISO of 100 (or 44), DR is 11.51 EV. At ISO 800 (or 360), DR is 8.56 EV, a loss of 3 stops.

If we look at the same progession for the 5D2, we have a DR of 11.16 EV at base ISO (73) and a DR of 10.66 EV at ISO 800 (or 564), a loss of just half a stop.

Ray, I must say I too was befuddled with this paragraph and also couldn't see (and still don't) the relevance of pixel count to FOV and DOF. You can have a sensor of any pixel count you want and the FOV and DOFwill still be determined by the focal length of the lens, the F-Stop and the lens to subject distance, whatever the resolving power of the sensor. Stitching DSLR images comes into play where you want higher resolution to cover the same FOV and you're not using a P65. And you'd need a longer lens to capture roughly the same FOV on a MFDB than on a DSLR with the same aspect ratio - if I recall from the film era when you needed an 80mm lens on a Rolleiflex to get roughly the FOV of a 50mm lens on a Leica (aspect ratio excepted). No?

Bernard, "funky" Bayer matrices have been serving us well, just to judge from the superb quality of your own results with your cameras embodying that technology, and not to start another battle - so far Foveon hasn't demonstrated any superiority - so I'm not sure I understand what this "funky-business" is. But how does one of these scanning backs deal with RGB interpretation? Is it Foveon-type technology? What do they do to capture different wavelengths of light and encode it as data to be interpreted as colour?

Granted, Bayer works very well. Now it has to be the case that the colors delivered by a true RGB device are significantly better but it is true that we do nothave a good metrics for this, nor evidence that our brain is able to actually perceive this difference at the conscious level.

It would be interesting to blind test this and see if people feel a difference between equal resolution images shot with a Betterlight and a back. My bet is that they would feel a difference, but be totally unable to tell us what the difference is.

The backs simply work by moving an array with 3 lines of sensors, filtered for RGB each in succession so that R, G and B pass in front of the same pixel scene on after the other. Obviously artifacts are introduced if there is the slighted movement of either the back of the objects in the scene. I am not saying that this is a practical solution for outdoor work...

Cheers,
Bernard

Yeah, well if you had a Thermos of cappuchino and an iPod full of Mahler symphonies you could sip coffee and listen to a symphony in the great outdoors while each shot processes...............

Anyhow, thanks for the insight on how they work. I can imagine it being ideal for indoor repro work of inanimate objects where the photog is paid by the hour.

Mark,
You seem to be confusing the role of pixel count with pixel density in this context. When emulating the result from a single P65+ shot, by stitching images from a smaller sensor, it's the pixel density of the smaller sensor that will determine the choice of appropriate focal length for both equal FOV and equal pixel count of the final stitch.

It is the pixel count of the smaller sensor that will determine the number of images required to be taken for stitching. Different size sensors of equal pixel density will of course have a different pixel count.

For example, the Canon 1Ds3 has the same pixel density as the cropped format 20D and almost the same pixel density as the P65+, therefore, whether I stitch with the 1Ds3 or 20D, I will use the same focal length of lens, which will also be approximately the same focal length as that used for the single P65 shot.

I will of course need to stitch a greater number of images using the 20D instead of the 1Ds3.

However, if I were to substitute the 15mp Canon 50D for the 8mp 20D, and use the same lens, I could certainly get the same FoV in the final stitch from the same number of images stitched in exacly the same way, but the final stitch would have almost twice the pixel count of the single P65 shot.

In order to achieve the goal of equal pixel count, which is the purpose of the stitching exercise, I would need to use a shorter focal length of lens with the 50D.

Of course, if I don't care what the pixel count of the final stitch will be because I intend to either upsample or downsample the stitched image according to print size, then I could use any focal length I liked. The longer the focal length, the more images I would need for stitching purposes and the greater the pixel count of the final stitch.

However, in this exercise we do care about pixel count, don't we?

Ray, not clear to me what I'm confusing..otherwise there would be clarity, eh? Do we understand the same things by these terms?

Pixel density and pixel count: Density is about how many pixels you cram into a given space. Count is the number of pixels. For a sensor of fixed dimensions, pixel density will be higher the higher the pixel count. A very large sensor with a large number of pixels could have higher or lower density than a DSLR with a smaller number of pixels. It all depends on pixel count relative to sensor size. BUT none of this pixel density and count as defined here has anything to do directly with FOV and DOF. Those are determined by sensor size, focal length and lens to subject distance.
Mark

I can't see the difficulty here, Mark. As Jeff Schewe has admitted, the reason one might need a P65 is to make big prints of high quality. If you want a 60mp stitch using a small sensor, say a Canon 20D, you will need to stitch a certain number of images, perhaps 12 allowing for overlap. You'll need a lens of around the same focal length as you would use with the P65 because the pixel density of the 20D is about the same as that of the P65+ (in fact it's slightly less).

If you use a 50D instead of a 20D, you'll need a shorter lens, even though the sensor is the same size, and consequently you need to stitch fewer images to get your 60mp final stitch with the same FoV. If you don't use a shorter lens, the final stitch will have a significantly higher pixel count. What's so difficult to understand?

However, it's true that in paractice, if I were to carry a 50D up Poon Hill instead of a P65+ with a view to stitching a few images to get a file size suitable for the same size print I'd be able to make from a single P65 shot using, say, a 24mm lens, I wouldn't be too worried about getting an even higher resolution image from the stitching process.

So lets say I use the same focal length of 24mm with the 50D that I would have used with the P65+ (or 20D) for the purpose of capturing the same scene from the same position. If I use the same F stop as I would have used with the P65+, I can expect to get a sharper result since system resolution is always a product of lens resolution and sensor resolution and the 50D, with its higher pixel density, should deliver higher resolution than the P65+ from the same quality of lens (excluding the issue of the AA filter. One would need to see real-world results to make a judgement on this).

If we assume that the 50D would be capable of higher resolution at the same f stop, then one can trade such higher resolution for greater DoF by stopping down, say, one stop.

However, I would be prepared to accept that the P65's advantage of a lack of AA filter may affect such theoretical predictions, if someone were to show me some real-world stitches demonstrating such a comparison.

My own tests comparing the 50D with the 40D have demonstrated that a 50D image at F11 has about equal resolution to the 40D at F8 (and I expect greater resolution than the 20D at F8 although I haven't done the comparison).

This is a factor which is often overlooked by those who complain that lenses are not good enough for cameras with ever increasing pixel count. When the 50D was announced, there was much discussion about the usefulness of such a high-pixel-density sensor. Perhaps no purpose would be served by stopping down beyond F8. Well, the purpose is increased DoF with no loss of resolution at the plane of focus, compared with a 40D or 20D.

The Betterlight backs take 35 to 110 seconds for full res, and the Seitz about a couple of seconds - fast enough to only slightly blur water spray... not fast enough to catch a vaulter in flight.

True, the Seitz is fast, but the image quality I have seen from it was disapointing, like in "far worse" than what the best DSLRs deliver because high ISO is needed to achieve these fast scan times.

There are obviously very few reports about the Seitz, and anybody with different experiences, please post samples.

Cheers,
Bernard

The solution would seem to be to have not three sensors for each row of pixels, not four, (with two green, 1 red and one blue) but 32, or 64, so you can add up the photons for each sensor/pixel for real high ISO.

I have only seen Seitz pictures on their stand at Phocus/NEC/UK, but they looked OK.

I'm surprised no one has mentioned Circle of Confusion (COC). Everyone seems to be leaving it out of their definitions of DOF.

FOV is controlled by the sensor dimensions. This means a smaller sensor camera using the same focal length will have a smaller FOV. By stitching, the smaller sensor camera can attain the same FOV as the larger sensor camera.

If the pixel density of both cameras is the same, therefore using the same COC, the image will be nearly identical. Usually the smaller sensor camera will have a higher pixel density, which means a smaller COC can be used and both DOF and resolution will be larger. Other factors that determine IQ may be more important than resolution.

This is something I have never considered before, since in practicality you do not purchase a camera based on it's stitching characteristics. After consideration technically Ray is correct on this one. Again the relevance still escapes me, and that doesn't excuse the fact that he made an incorrect statement trying to claim a depth of field advantage of stitching 5d Mark2 files over using a p65+ back, and then tried to pass that off as some kind of test, instead of simply admitting he didn't think it through (which we have all done) when he was called on it, because the depth of field advantage of the 5d mark 2 is insignificant.

If two cameras have exactly the same pixel density, and yet different size sensors, to create an exact match in pixel size and field of view will require the exact same focal length lens. Granted, when stitching one nearly always uses a telephoto lens because the goal is putting the detail of the scene on more pixels when stitching. I'm not sure who would ever worry about this, because when you decide to stitch you are doing so because the camera you have ... the one you purchased based on your want/need/afford ... doesn't have the resolution to deliver the quality you want when you are printing the file later on.

For me it was easier to understand thinking of it from a little different perspective. Imagine you have a 4x5 view camera, and for that camera you have two backs, one a 4x5 back that is 8000x10000 pixels in resolution, and the other is a 2x2.5 inch back that is 4000x5000 pixels in resolution. Both backs have exactly the same pixel density, one is just physically smaller. If you were to take an image with the larger back, and try and duplicate it with the smaller back, you would simply place the smaller back in each corner and then stitch the four captures into one single capture. The two captures would be identical in pixel dimensions and FoV, as well as depth of field. No change in focal length.

If the smaller back had a pixel density greater than the larger back, to match the exact FoV and pixel density of the file would require a slightly wider focal length lens, because the back is capturing more data per projected area than the original back. This would result in a DoF improvements, although for this to be significant the pixel density would have to be significantly different.

As I said the relevance escapes me ... it's sort of like something I'd expect in a trivial pursuit game.

When you pan and stitch, you loose a lot of pixels through having to crop back to a rectangle.

To make good use of a higher density sensor (smaller pixels), you need to use a smaller COC, but with smaller pixels and higher density sensors the magnification (reproduction ratio) is lower (smaller image size), and the DOF is higher. (I think DOF depends on COC magnification and aperture), as explained on page 33 of Harold Merklinger's "the ins and outs of Focus" (which was downloadable foc).

A smaller COC gives you a smaller DOF, but (for the same number of pixels) a smaller pixel size gives you a smaller image, a smaller magnification and a bigger DOF.

Wayne and Ray, OK, the way Wayne has related these things it now makes technical sense, but so what? If Ray's basic argument is that you don't need to spend 40K to get a high density capture of a large scene because you can stitch - that's true as far as it goes. But it may not go all the way, so to speak, because there are other unaccounted factors which differentiate IQ between these technologies. I've seen a large panorama from a suite of P65 images and it's mind-blowing. I believe there are more factors distinguishing these trechnologies than what we are discussing in this thread; not sure exactly what, but I'll be learning more about that. On the issue of Dof and CoC, I wonder whether diffraction (as a function of f/stop) affects these sensors differently.

Okay! I see Im gonna have to save face .

You do seem to be rather sensitive about this issue, Wayne. I did make the following general statement about DoF, in post #102 repeated below, without specifically mentioning the 5D2 or stitching. However, because the previous discussion had referred to full frame DSLRs such as the D3X and 5D2 being used for stitching, it would be reasonable for anyone reading that statement to assume that my general statement included the 5D2. In that sense the statement was at least partly incorrect and certainly misleading.


"There also seem to be some serious disadvantages to the MFDB system, such as a slow continuous frame rate, (less than one frame per sec with the P65+), reduced performance at higher ISOs (or reduced pixel count as an alternative) and shallower DOF at any given F stop (for same FOV), which is not always ideal for landscapes. (Not to mention poor autofocussing).

For example, in the above panorama of the Himalayas, how would you get both the ladies and the mountains sharp, using a P65+? F22 at ISO 44 in the early hours of the morning?"

So apologies to anyone who rushed out immediately and bought an inappropriate camera for the job, as a result of my misleading statement.


The truth of the matter is, I realised immediately after I'd posted that general statement that it was misleading and that perhaps I should go back and modify the statement. I didn't because I was a bit slack, and also I really did genuinely wonder (God's honest truth!) if anyone would pick that up. You did, but for the wrong reasons.

The last sentence in my statement in post #102, "For example, in the above panorama of the Himalayas, how would you get both the ladies and the mountains sharp, using a P65+? F22 at ISO 44 in the early hours of the morning?" could also be misleading, again implying that the 5D2 at F22 would produce greater DoF. But notice I didn't specifically mention DoF here. I used the word 'sharp'.

Having direct personal experience of those conditions up the mountain, I can confirm that at F22 and ISO 44 the heaving bodies of those ladies in the foreground and the swaying grass in the early morning breeze would not have been sharp due to the very slow shutter speed of the P65+ at base ISO. Here again is where the 5D2 would have an image quality advantage. If the depth of field desired required the use of F22 but the shutter speed desired in order to freeze the foreground required the use of ISO 400 or even 800, then the 5D2 stitch would likely produce the better quality image, in terms of DR, tonal range etc.

Now, you may think this is all irrelevant and merely a trivial pursuit, Wayne, but I consider it all necessary information in order to select the best tool for the job.

No pixels are lost during stitching Depending on the equipment used, there may be overlapping pixels of which you get to choose which to use in the final image. Stitching software only needs to know the pitch and yaw of a frame to know where to place it. If you're using a system with very accurate rotations, you can specify where to place the frame with little or no overlap. Stitching software like AutoPano Pro and Microsoft ICE don't have user settable pitch and yaw, so require more overlap.



COC is something you use to predetermine (or describe) what the DOF and/or diffraction will be for an image at a particular f stop. It is not settable. Higher pixel density sensors will give you more resolution for the same lens (Big Assumption: the lens can out resolve the sensor). In practice, the same lens isn't used on both cameras and generally speaking the lens for the smaller sensor camera will be better corner to corner than a comparable lens on a larger format camera (Yes, there are exceptions to every rule).

I think it's safe to say a stitched image can easily match or exceed the resolution of a larger format camera. For example, an image from my Olympus e330 will have nearly 110 megapixels when stitched to the same size as P65+. What isn't clear is, will the IQ of the image be better? If not what is the criteria used to judge the IQ?

Ray,

At f/22 you would surely have diffraction issues with the 5DMK2 no? One of the more recent fads making the PS tutorial circuits these days - and quite a neat play actually - is to do "DoF blending" where you use less diffracting f/stops, and - in one variant of the technique - make two images - one focused for the foreground, another for the background, then blend them with a gradient or two in PS. This kind of technique would be just a bit trickier to implement in a stitching context - more images to deal with, but in principle may be something to think about when you go up the hillside again with your next batch of pretty ladies.

Mark