MP Equiv to Film Sq Inches: Why so Non-Linear?

[Jack Flesher]

P. S. Large format enthusiasts should probably stick mostly to the legitimate claims of
1) greater print resolution in lp/mm on equal sized prints, after the lower degree of enlargment needed
and
2) finer tonal gradations/greater dynamic range
to explain this preference. Those two have always been and probably always will be the main real reasons to sometimes choose a larger format than most others photographers use.
[a href=\"index.php?act=findpost&pid=66554\"][{POST_SNAPBACK}][/a]

And 3) Easier scheimpflug adjustments due to the longer focal lengths used at any given magnification ratio; 3a) And thus in many (but not all) compositions the appearance of greater DoF compared to smaller formats.

[Mark D Segal]

Hi Gary,

Since I'm using an 8.3 megapixel EOS-20D, which has fewer pixels and a smaller sensor than any of the cameras in the test, I couldn't make that statement myself. But now that you've made it, I'm happy to concur. 8x11 prints from the 20D are incredibly fine; it's hard to imagine better. So yes, there's no point at all in having more pixels (read, higher cost) unless you plan to print considerably larger. I suspect all the participants have Epson 7800 or 9800 printers. By the way, a friend from HP tells me they're coming out with some large format printers later this year that will give Epson a real run for it's money. I can hardly wait. I love big, fine prints.
[a href=\"index.php?act=findpost&pid=65995\"][{POST_SNAPBACK}][/a]

Norm, the point of having higher MP count whatever your print size is that the more MP you have, the more cropping you can do while preserving high quality in large prints of cropped images. It just gives added flexibility for all those who may see the need to improve composition by cropping their images in post-processing. How much that is worth varies of course from one photographer to another.

[Gary Ferguson]

I'd agree that larger image circles generally come at the cost of reduced resolution. But in the real world of available optics there's some notable examples that manage to depart from a smooth linear relationship.

Rodenstock's Apo Sironar Digital HR lenses, with image circles of about 70mm, manage to deliver the kind of performance I'd normally associate with excellent 35mm lenses with their typical image circles of about 45mm. And there's others, Hasselblad's V system 40mm IF, or the Contax 645 120mm Macro, or the new Sinar/Zeiss optics, or the Canon 300mm 2.8L compared with the equivalent optic for a smaller format than 35mm.

So even though I accept the general thesis of the trade off between image circle and resolution, there's sufficient spectacular exceptions to the general rule so that individual lenses still need weighing on their own merits.

As for camera movements coming to the smaller formats, yes that's self evidently so. However, in my experience of using a 37x49mm digital back on a Linhof, and also of using Canon's tilt and shift lens range, I'm not completely convinced that movements are quite such a happy marriage as with the more traditional formats such as 4x5.

The problem IMO is achieving the increasingly exacting movement precision required as sensor or film formats shrink. One simple example is loupe magnification. The x4 loupe I used with 4x5 was consistent with my typical print enlargement, but if I try and use a x7 or x10 loupe on a 37x49mm sensor then the image is simply lost in the grain of the ground glass.

[BJL]

And 3) Easier scheimpflug adjustments due to the longer focal lengths used at any given magnification ratio; 3a) And thus in many (but not all) compositions the appearance of greater DoF compared to smaller formats.
[a href=\"index.php?act=findpost&pid=66563\"][{POST_SNAPBACK}][/a]

Ah yes, I did mention that earlier in my post. Though strictly speaking that is an advantage of a view camera, regardless of format size itself.

As Gary mentions, using smaller format sensors with view cameras seems to bit of a hack at least for now. I wonder if anyone will produce a view camera system designed from the ground up for use with smaller sensors, say via digital backs with 24x36mm or 36x48mm sensors. I understand that "medium format" style sensors without microlenses are best for this, so strapping a DSLR onto the back of a view camera might not be the ideal solution.

[BJL]

I'd agree that larger image circles generally come at the cost of reduced resolution. But in the real world of available optics there's some notable examples that manage to depart from a smooth linear relationship.

Rodenstock's Apo Sironar Digital HR lenses, with image circles of about 70mm, manage to deliver the kind of performance I'd normally associate with excellent 35mm lenses with their typical image circles of about 45mm.
[a href=\"index.php?act=findpost&pid=66573\"][{POST_SNAPBACK}][/a]

The new digitally oriented Rodenstock and Schneider Digitar lenses do seem to push resolution forward in "medium format" sizes, and a rough comparison of MTF for the Digitar's at 40l/mm to data for Canon lenses at 30lp/mm does suggest that these new "digital large format" lenses have comparable or better resolution than good Canon primes.

But perhaps the more interesting comparison is to equally new high resolution lenses for smaller digital formats. Canon has talked about producing new sharper lenses for the needs of high resolution 35mm format digital, and hopefully will publish MTF data for them at higher thn 30lp/mm. But for now the best comparison available to me is 60lp/mm MTF data for Olympus Digital Zuiko lenses and for Schneider Digitars. (Are MTF graphs availablefor those Rodenstocks?)

Again, all four of the Zuiko Digital non fish-eye primes have 60lp/mm MTF far higher than any Digitar, and similar to or higher than the 40lp/mm of any Digitar. If you look towards the corners, the 60lp/mm MTF of the Zuiko Digital primes is similar to even the 20lp/mm of most or all Digitars.

[neal shields]

One suggestion that I would have to make the tests more defendable and easier to compair would be to either use a test target that had deminishing resolution or simply back away progressively by doubling the distance each time.

Then if two lines on a bill could be resolved with all systems at 10 feet but only selective system at 20' this gives a quantifiable difference that isn't dependent on opinions.

The FBI spent about a year compairing systems and concluded that 35mm film at 200 asa is about 16 meg.  I can't understand why there is so much difference between their test results and some others unless it is method.

http://www.fbi.gov/hq/lab/fsc/backissu/apr...swgitfield1.htm

It seems to me that unless you are developing quantifiable objective numbers with you testing that are independent of subjective opinions you are simply compairing initial outputs of one system to another.

Film will always lose as it is orange and reversed in colors.

I have also found in my own testing, using charts and microscopes to read the negatives that fill captures about 130% the detail of a 9 meg camera.  That comes close to agreeing with the FBI.

I might add that in my tests, I can get way more detail onto film than can be seen with less than a good 8000 dpi scan.  So some people also fall into the trap of compairing direct capture to scans, not to film.

For many people the work flow is the most important thing.  Which camera will get me where I want to go quickest and easiest.  Digital usually wins.  However that is a different question than: "which camera captures the most information?"

The reason I feel the second question is the most important one is that with modern computers and programs I can synthise almost any "look" I want.  I can certainly make a low res picture "look" "sharper" than a high res one.  

About the only thing I can't do is create information that the camera didn't capture.

[Jack Flesher]

The FBI spent about a year compairing systems and concluded that 35mm film at 200 asa is about 16 meg.  I can't understand why there is so much difference between their test results and some others unless it is method.

http://www.fbi.gov/hq/lab/fsc/backissu/apr...swgitfield1.htm

[a href=\"index.php?act=findpost&pid=66653\"][{POST_SNAPBACK}][/a]

Maybe because they did that test about 5 years ago and there have been significant improvements in digital sensors, camera firmware and raw converters since then?

[Gary Ferguson]

The FBI spent about a year compairing systems and concluded that 35mm film at 200 asa is about 16 meg.

Neal, a few points.

It wasn't "the FBI" that conducted the test, it was an individual within a department of the FBI, so we shouldn't give the results more prestige than they warrant. If the same individual had been working for the Kazakstan Dental Association would we be quite so respectful of the results?  

There was a debate recently on the Leica forum at Photo Net regarding photographs as evidence. It wasn't conclusive but there were clearly experienced professionals using photographs around the world in courts of law who believed film was the better law enforcement choice. Not becuase of resolution but because a negative is more likely to be accepted as untampered and authentic than a digital file.

And finally, as has been pointed out, that report is ancient history. If I remember correctly (and I'm sure you'll bring me to book if I'm wrong) the 16MP conclusion was extrapolated, they didn't actually look at a 16MP file, they just guessed.

Now none of this makes their conclusions wrong in itself, but we should see it as just another data point in a rich and complicated debate.

But really, who cares? I use both film and digital and I rarely choose one rather than another because of quality. With a bit of application I can usually get the quality I require from either medium at the print sizes I commonly use. The choice is more governed by look and application than lppm.

[neal shields]

Neal, a few points.

It wasn't "the FBI" that conducted the test, it was an individual within a department of the FBI, so we shouldn't give the results more prestige than they warrant. If the same individual had been working for the Kazakstan Dental Association would we be quite so respectful of the results? 

There was a debate recently on the Leica forum at Photo Net regarding photographs as evidence. It wasn't conclusive but there were clearly experienced professionals using photographs around the world in courts of law who believed film was the better law enforcement choice. Not becuase of resolution but because a negative is more likely to be accepted as untampered and authentic than a digital file.

And finally, as has been pointed out, that report is ancient history. If I remember correctly (and I'm sure you'll bring me to book if I'm wrong) the 16MP conclusion was extrapolated, they didn't actually look at a 16MP file, they just guessed.

Now none of this makes their conclusions wrong in itself, but we should see it as just another data point in a rich and complicated debate.

But really, who cares? I use both film and digital and I rarely choose one rather than another because of quality. With a bit of application I can usually get the quality I require from either medium at the print sizes I commonly use. The choice is more governed by look and application than lppm.
[a href=\"index.php?act=findpost&pid=66664\"][{POST_SNAPBACK}][/a]

[neal shields]

I thought he had done the tests as part of his work, but in any case it is published on the FBI's web site and looks like an official recomendation to me.

Also, I don't think that there are many people saying that a 2006 pixel has more resolution than a 2000 pixel.  Larger pixels have more dynimic range than small ones but pixels are getting smaller not larger in most new cameras.

Beyond that this gets into a debate about what is "science".  In the book by the Nobel prize winner who basically made gene maping possiable, (Dancing Nakid in the Mine Field) he talks about the fact that science is: coming up with a theory, runing experiements to prove the theory and publishing the results in a manor that other people can check, by running the same experiements and seeing if they get the same results.

Yes I can order the disks and get the raw files to see which image I like better but that doesn't tell me much about absolute resolution limits.  To make judgements about resolution I need to have some deminishing detail so that I can say that I can read this line but not that line, for example.

To do that you have to quantify something.  The title of the article is "measuring megapixels".  Normally if you "measure" something you get numbers, not opinions.

People have been testing resolution for over a hundred years.  Why not use the proven techniques?  (You can download a resolution test chart off the web.)  

Yes I know, people will say, "I don't photograph resolution test charts".  Point is, if you are trying to decide what you like best for your work, the test is a great test.  Just don't advertise it as a resolution test.

As to who cares, I assumed that was the purpose of the article.  If resolution and absolute information capture isn't important to someone's efforts or art then that is fine.  I suspect that for 99% of the world's photography (a vast amount better than mine) resolution isn't a factor.

One of the reasons that I believe the FBI article is that it tracks my own tests.  However, I collect microscopes so I am checking my negatives under the microscope, not by using low res scans.  With very carefully made 4x5 negatives, I found that I had to have an 8000 dpi drum scan (wet) to get all the information off of a velva transparancy.

The easiest resolution test that I can come up with is to photograph a newspaper and keep increasing the distance.  Different camera systems will drop out as you try to read the paper from the file or negative.  However, you will have to have some very high resolution way of checking your negatives. (A 40 power loupe, a microscope or a very high res scanner.)

[BJL]

The easiest resolution test that I can come up with is to photograph a newspaper and keep increasing the distance.  Different camera systems will drop out as you try to read the paper from the file or negative.
[a href=\"index.php?act=findpost&pid=66721\"][{POST_SNAPBACK}][/a]

That would measure about the same thing as high contrast test pattern images, still rather removed from the more important issues of photographic fidelity relating to reproducing detail at boudaries of far lower local contrast, and reproducing contrast levels in the final image that are reaasonbaly close to the original subject matter. This is very different from turning a "black and white" subject into a "49% gray and 51% gray" image on the print, even if you or the FBI can still just barely read the writing on the "gray and gray" print.

Referring again to Fuji's datasheet for Velvia: it cites 160lp/mm for an extremely high contrast (1000:1) test pattern, but only 80 lp/mm for a low contrast test pattern, and I have read skeptical comments from film reiewers about even the latter lower figure. No film reviews I read put much weight on the high contrast test pattern number, but of course it was always the one that got most prominance in film advertsing, with some film makers publishing only that number, and of course it is still the favorite of the film rear-guard.

The difference between those two numbers (160, 80) already gives a factor of four difference in "pixel count equivalent", showing how much room there is for fudging the answers by choosing a measurement that favors the result one wants to get.


My preferred style of test: take some real photographs of interesting, finely detailed subjects, and view them (not images of newspapers or test patterns) from various distances.