Comparison of 4 film formats and 2 digital cameras

[Bert Otten]

Recently I finished some experiments on the image quality and "real-life" resolution of four film formats (8x10, 4x5, 6x7cm and 35 mm) and 2 digital cameras (6.3 and 5 megapixel). I've given the results of these experiments on my website. After having gone through a first round of improvements, I'm ready for the next one. Please give me your comments. Thanks.

Site with
experimental results (http://www.oprit.rug.nl/otten/Comparison.html)

[Graeme Hird]

Bert,

You've confirmed my thoughts, but I think you might be preaching to the converted here. Still, it is nice to have a real world comparison to refer to.

Can you do another out in the field somewhere?

Cheers, Graeme

[Ken Lee]

Excellent !

[David A. Goldfarb]

Good test. Among other things, it resolves my dilemma of whether it would be better to invest in a 5400 dpi scanner for 35mm or in a 6.3 mpix digital body.

[Ernest Purdum]

I appreciate the amount of work put into this study. My thanks. Some of your comments go over my head because of my ignorance of digital matters. One comment brings up a question, the mention of film advantage in extreme heat and cold conditions. I have been thinking of carrying a digital camera in my car where it would be exposed to more heat than film would tolerate. Is this a bad idea?

[Leonard Evens]

A very interesting study. I've seen limited results about parts of the picture, including my own, which these results confirm. But you've done a thorough analysis and put it all in one place. Great job!

One comment. You mention at one point that the digital camera had a wider exposure latitude than scanning film. One can do even better by scanning negative film, color or b/w. The use of reversal film has become ingrained because editors and other consumers of professional photography have found it convenient. But if you are going to scan, to my mind, the disadvantages of reversal film outweigh its advantages. It is true that you have the slide to compare for color balance, but there is no special reason that what you see in the slide in a truly accurate rendition of the color in the scene. Using gray cards or just neutral parts of the scene, one can usually do just as well or better with color negative film.

[bob moulton]

Carefully done and well detailed results. One question: Do you think the use of better digital equipment would alter the results? The bodies and lenses on all the film cameras are probably better than those of the digital cameras. One wonders what would happen if you used a 6.0 or above megapixel DSLR with an OEM lens. And since you include a wide array of film cameras, one also wonders what the results would be if you had the opportunity to include digital results gathered from LF digital backs and from the current crop of 14 MP digital cameras as well as from the hybrid recently released by Leitz.

Of course performing those experiments means you must hit the lottery!

[George Stewart]

Excellent! Just what I wanted to see. I liked the fact that you used a scanner that most of us can afford. But I'd still like to see the results from pro-level scanners. I also like the idea of adding some outdoor photos. Thanks for your efforts.

George

[Archie O. Alcantara]

This was an excellent study. As the results from film were scanned, I wonder how this would compare if the film pictures were processed the old fashion way via darkroom ?? I understand that the results could still be constrained by the scanning process of the print to post to the web, but I guess my curiousity is how these result will translate to a printed medium.

Again thanks for sharing this information !!

[Ellis Vener]

I looked at his comparisons, and i don't think much of them.

with regard to the film tests the problem is the scanner. with the dslr the problem is the lens, and he doesn't say whether he was shooting raw files or jpegs; or how the scans and digital photos were processed in photoshop, I'd love to see the original digital files.

He then takes this flawed basis -- he assumes that the quality level of the 300D is representative of all digital cameras (it isn't) -- and extrapolates flawed conclusions from there.

in sum: with film he is actually evaluating the scanner, and with the DSLR he uses a lens that isn't equal to the quality of the lenses used on the film camera, etc.

his conclusions are flawed

[Jorge Gasteazoro]

He then takes this flawed basis -- he assumes that the quality level of the 300D is representative of all digital cameras (it isn't) -- and extrapolates flawed conclusions from there.



Correct me if I am wrong, but isn't the 300D a middle of the road camera most likely to be used by a non pro? Isnt this camera with the lens that comes with it an equivalent of the Canon 35 film camera?



I find this objection curios, because with a 35 mm film camera the body does not matter, sure some bodies can do double exposure etc, but the body is in most cases unimportant. Are you telling me that if I want to take digital pictures that can equal the quality of a run of the mill 35 mm slr I have to go and buy special lenses and a top of the line pro camera? This does not make sense to me.

[Richard Fenner]

May I inject a small note of criticism? I have a horrible time when it comes to using numbers in different contexts - you should see me shake when I use stats packages! But, it seems to me, when I look at the bottle shots, taking the 6x7 image, I'm looking at it as if it was taken from a print approximately 3m by 3.5m (a bit shorter, as 6x7 isn't 6cm by 7cm). For the 8x10 and 4x5, as if they were 3.12m by 2.5m (the changed degree of enlargement in the image presented accounts the format differences, so they're effectively the same size). Previous comments on this site indicate a MAXIMUM normal print size for most people to be about 1.25m by 1m (50x40) - 40% of the enlargement shown. It would seem to me, to be meaningful at a print size of 50x40, the images shown would have to be reduced by 60%. For the people who typically make even smaller enlargements, (and from recent threads, this is the majority), an even greater reduction would be required.

In short, 50x is predisposing the results to be those we would like to see - the greater the magnification, the larger formats must by definition win, regardless of the real-world ability to discern a difference in prints on the wall (which I take to be 'real world' - others may have different criteria).

Or have I blundered with the numbers?

Richard.

[Bill_1856]

I do not believe that the results, as published, represent an accurate comparison of the best (or even the usual) results obtainable from these diverse media.

[Darin Cozine]

Yes there are a few problems with this comparison.

The LF equipment was tested with top-of-the line optics. Meanwhile the digital cameras and lenses were entry-level. I have seen another test which compared a Canon EOS 1d 12mp with its counterpart EOS-1n. The same L-series lens was used, and various films were tested. The film was scanned with various methods and a microscope with a digital camera.

The plaubel Makina has performed poorly in other tests due to a misaligned rangefinder. A Mamiya 7 or Hasselblad would perform much better.

[Bert Otten]

To give some short answers:

Ellis,

if you see the captions under the USAF charts: raw files were used of the digital cameras, so I did state that.

I used the 300D as a representative of a good 6.3 Megapixel camera and showed that there is quite a big difference with a 5 Megixel compact. I did not extrapolate anything to other digital cameras: If you look in my Printsizes part, I put question marks behind my calculations of a 14 megapixel digital camera, because I can not be sure.

I did include the data read from the film directly before scanning (indicated by “original” in the tables), so I have not only tested the scanner, although I did state that the scanner was the weakest link. Today I have looked with a microscope at the air image of the plastic lens of the Canon (18-55) at the same focal length setting as in the test and seen that on axis the image of the USAF chart is resolved with the finest detail blocks at full opening of the lens. This implies that the lens has a resolution of a factor of at least 2.5 times higher than the test result of the whole camera. The lens was not limiting.

Ellis, I think therefore my testresults are useful, since I have answered all your points of criticism. Thanks for your input.

Richard,

Yes, the format of the film largely determines the printsize possible that is still sharp to the human eye at reading distance. The beer bottles represent different amounts of magnification, indicated under the images. The point is: If I would like to print the presented scene from all cameras and formats on the same size, what can I expect? Finally, my table of maximal printsizes is a practical upper limit that is handy when you prepare an exhibition

Bill,

It is hard to answer criticism that contains no arguments. “Believing” is not part of experimental science, only a start of doing experiments.

Darin,

I have aligned the rangefinder of the Plaubel Makina, and if you look at the numbers, the film shows a resolution of 55 linepairs/mm in the filmplane, which is the same as the results in the other large cameras. Given the time, I will repeat the experiment with a Hasselblad, but I don’t expect wildly different results.

Thanks very much for your time!

[Ellis Vener]

I find this objection curios, because with a 35 mm film camera the body does not matter, sure some bodies can do double exposure etc, but the body is in most cases unimportant. Are you telling me that if I want to take digital pictures that can equal the quality of a run of the mill 35 mm slr I have to go and buy special lenses and a top of the line pro camera? This does not make sense to me.

let's use the automobile as a metaphor . The engine and the chassis of a digital camera are the sensor array that converts the photons that strike it nto sanalog signals which are then processed into a digital signal (zeros and ones). The camera body as you are thinking of it -- chassis, meter, lens mount, shutter, etc. is mostly irrelevant to the extent that they don't interfere with the quality of that process. You can put a 250 hp engine and aterific transmission in any car and it will go like a bat out of hell (for awhile). Similarly you can but a Yugo engine and transmission in a Corvette or BMW and turn the car into a chump. You can also take a good sensor array (the CMOS or CCD "chip") and then use a variety of firmware & software programming, as well as processing hardware to get different kinds of results. Sort of like mixing a BMW engine, and then your choice of transmission. So to be more precise, when I say "digital camera" I am referring to the quality of the camera's digital drive train , all aspects of which to a greater or lesser degree have an effect the quality of the image.

Lens quality is an undeniable factor in the quality of image making whether we are discussing digital or film based photography. In general digital cameras do need higher quality lenses than film based imaging requires. This is because ofthe different mechanical nature of the two types of medium. Put a medium quality of lens on the front of a digital camera and you limit the quality of the output. The immutatable rule of computation is "garbage in, garbage out."

With digital imaging you are not justthe photographer but you are the lab too. if you choose to use a "raw' capture mode you are also to a large degree the "film" manufacturer as well. And atthat point it comes down to how you process the image and the tools you use to do the processing with. Any version of Phase One's Capture One DSLR "raw conversion" software is more powerful than Canon's (as is Adobe Photoshop CS); even more powerful is the full version of Capture One (and in some areas it is even more powerful than Adobe Photoshop CS). The strength and quality of a digital imaging program are the algorithms used to translate the digtal signal back int oa visual form.

Otten does nt specify how the digital images --scans or direct in camera capture - were processed. 8 bit? 16 bit? Jpeg? Raw? Tiff? What software was used? what kind of sharpening was used (all digital images require sharpening), etc. , so basically all Mr. Ott is doing here is comparing results from a low to mid quality desktop scanner against the results of a low to moderate quality digital SLR against a low quality point and shoot camera. For all of these reasons, these tests really don't prove anything except that 8x10 & 4x5 film have highr resolution that smaller formats.

But more power to Mr. Ott for actually doing some sort of testing himself as opposed to just accepting other people's words at face value. He deserves (and gets from me at least) a lot of respect for doing that.

[Ellis Vener]

I find this objection curios, because with a 35 mm film camera the body does not matter, sure some bodies can do double exposure etc, but the body is in most cases unimportant. Are you telling me that if I want to take digital pictures that can equal the quality of a run of the mill 35 mm slr I have to go and buy special lenses and a top of the line pro camera? This does not make sense to me.

let's use the automobile as a metaphor . The engine and the chassis of a digital camera are the sensor array that converts the photons that strike it nto sanalog signals which are then processed into a digital signal (zeros and ones). The camera body as you are thinking of it -- chassis, meter, lens mount, shutter, etc. is mostly irrelevant to the extent that they don't interfere with the quality of that process. You can put a 250 hp engine and aterific transmission in any car and it will go like a bat out of hell (for awhile). Similarly you can but a Yugo engine and transmission in a Corvette or BMW and turn the car into a chump. You can also take a good sensor array (the CMOS or CCD "chip") and then use a variety of firmware & software programming, as well as processing hardware to get different kinds of results. Sort of like mixing a BMW engine, and then your choice of transmission. So to be more precise, when I say "digital camera" I am referring to the quality of the camera's digital drive train , all aspects of which to a greater or lesser degree have an effect the quality of the image.

Lens quality is an undeniable factor in the quality of image making whether we are discussing digital or film based photography. In general digital cameras do need higher quality lenses than film based imaging requires. This is because ofthe different mechanical nature of the two types of medium. Put a medium quality of lens on the front of a digital camera and you limit the quality of the output. The immutatable rule of computation is "garbage in, garbage out."

With digital imaging you are not justthe photographer but you are the lab too. if you choose to use a "raw' capture mode you are also to a large degree the "film" manufacturer as well. And atthat point it comes down to how you process the image and the tools you use to do the processing with. Any version of Phase One's Capture One DSLR "raw conversion" software is more powerful than Canon's (as is Adobe Photoshop CS); even more powerful is the full version of Capture One (and in some areas it is even more powerful than Adobe Photoshop CS). The strength and quality of a digital imaging program are the algorithms used to translate the digtal signal back int oa visual form.

Otten does nt specify how the digital images --scans or direct in camera capture - were processed. 8 bit? 16 bit? Jpeg? Raw? Tiff? What software was used? what kind of sharpening was used (all digital images require sharpening), etc. , so basically all Mr. Ott is doing here is comparing results from a low to mid quality desktop scanner against the results of a low to moderate quality digital SLR against a low quality point and shoot camera. For all of these reasons, these tests really don't prove anything except that 8x10 & 4x5 film have highr resolution that smaller formats.

But more power to Bert Otten for actually doing some sort of testing himself as opposed to just accepting other people's words at face value. He deserves (and certainly gets from me) a lot of respect for doing that.

[Ellis Vener]

I was busy responding to Jorge while Bert was responding. I double posted because I realized I was spelling "otten" Ott in at least (and hopefully only) one place.

[Bert Otten]

Ellis,

that's a good point: I only indicated that I used RAW as file format for the digital cameras, but not how I processed them. Neither did I include information on the color depth of the scans or sharpening. I will add that to the site. For now I can say that I did all I could to get the maximal resolution out of the digital files, so 16 bit, the Camera plug in of PhotoShop CS, in which I used the chromatic aberration correction option, but not the noise reduction. For the scans I only sharpened in PhotoShop at the 1 pixel level to 100%, to avoid any unwanted overmodulation, but to be shure I got the limit of resolution. So thanks Ellis, I will add the information to the site.

[John Kasaian]

Very enlightening! I won't be trading my 'dorff for an elph(but the thought never really crossed my mind anyway!)

[Brian Ellis]

I too applaud Bert for taking the time to do this and for sharing the results but I don't think this is an "experiment on image quality and real life resolution." What Bert has done is similar to taking six negatives made with six different films and six different cameras to six different labs and then using the resulting prints as the basis for comparing the quality of the films. I think we all would say "wait a minute, that's no way to compare films, things have to remain more constant than that from one film to the other if you're going to validly compare them." There are a huge number of variables here that affect the results including the equipment and methodology used to scan the negatives, the software and methodology used to convert the RAW digital files, the quality of the various lenses, Bert's ability to identically focus six different cameras six different times, etc. etc. almost ad infinitum.

This isn't intended as a criticism of Bert at all, like others have said I think it's great that he's doing his own "testing" and has taken the time to share the results. I just don't think this "test" should be taken as saying anything more than that these happen to be the results Bert obtained when he used the various cameras, lenses, scanners, software, films, chips, computer, monitor, etc. and that changing anything in the chain could and al ost certainly would change the results. I do know for a fact that the differences in the prints I make from my 6x7, 4x5, 8x10 and digital cameras are far far less than the differences shown here among these same formats.

[Bert Otten]

Thanks Brian for your response. Let me set one thing straight: I developed all 4 formats films myself, they were all Provia 100F and they were developed with a computerized machine, the Jobo ATL 2000 with very tight temperature and timing tolerance. So there are a few things constant here. The focus of the two large format cameras has been checked carefully on reproducebility. The focus of the Plaubel Makina has been checked by photographing a ruler, using the rangefinder, and has been corrected until precise. The focus of the Leica was very accurate, besides at the aperture used there is quite some tolerance. As I have explained on the site, I have done all I could to get the maximum resolution out of the RAW files of the digital cameras, and I have shown that the lens of the 300D was not limiting.

If you find that the difference between the prints from various formats is much smaller than the results I show, that can easily be explained by the limits in the printing process. You really have to produce very large prints to get everything out of an 8x10 transparency. So your statement is correct for prints of 40 cm wide and smaller, but not for larger prints.

Now here's my statement, and I'm very interested if you can falsify it by something stronger than impressions: The differences in image content I show are very robust as long as the quality of the optics is high, something like Provia is used and the focussing is correct. If you can show that you can get twice the resolution out of a 6.3 Mpixel camera as I did, I would already be convinced. Now to mess up a large format print so that you get much less resolution than I did is easy ofcourse, so that would not convince me.

My original transparencies are very close to the maximal resolution given by Fuji in their data sheet of Provia 100F. In fact I get close to that resolution in all film formats (75, 71, 63 and 61 linepairs/mm for the formats 35mm, 6x7, 4x5 and 8x10 respectively). As I have indicated, concerning the scans you look at on my site, the scanner was the weakest link. The optics and focussing were not limiting.

If you doubt my tests because your experience tells you something else, that's understandable, but only convincing if you have experimental results yourself!

Perhaps this is a good moment to ask if anyone on this forum is willing to repeat my experiment on any of the formats (including the digital ones) I have used. In science this is done very often: to try to repeat other ones experiments, to see whether they hold. Let us know.

Once again Brian, thanks for your reply,

[Brian Ellis]

Hi Bert - Thanks for your reply. I don't want to get into an argument, especially since I do (as I said before) admire you for doing the testing and posting the results. Just a couple points of clarification. When I mentioned focus as a variable I didn't mean the ability of the various cameras to focus accurately, I meant the ability of any human being to focus consistently and accurately. I seem to remember somewhere seeing the results of focusing tests showing that even the best photographers could only focus on the same point something like five out of ten tries or something like that.

With respect to print sizes, you're absolutely correct. I didn't mean that just because I see relatively little difference in my prints (which usually are no larger than 11x14) among the four different formats we both use, no differences ever exist in any prints of any size. I just meant that in my "real world" of making prints (as opposed to enlarging to the point you did in your tests) I don't see the kinds of differences your tests show.

[Bert Otten]

Good Brian, thanks for replying to the issue of printsize.

Concerning the issue of focussing: my standard deviation of focus position of the lens over a series of 20 times focussing on the 4x5 camera is 0.1 mm. Even if we use an extremely critical circle of confusion of 0.01 mm, the depth of focus is 0.33 mm for the used lens of 150 mm and f16 and the object distance used of 4 meter. That means that focussing was not limiting, since my accuracy of focussing was well within the depth of focus.

The only issue that could have infuenced sharpness of the testshots in the large format cameras is shift of focus due to spherical aberration of the lens, since I focus at f5.6 and shoot at f16. So I did a number of focus experiments with various apertures using the Rodenstock Sironar N lens. It appears that the standard deviation of focus position of the lens increases proportionally to the f number, but the average position of focus does not change (statistical t-test used). Hopefully these data gives you peace of mind concerning the accuracy of focussing in my experiments.

[Bert Otten]

A small clarification of the data I gave:

The standard deviation of focussing of 150 mm focal length at f16 is 0.1 mm The standard deviation of focussing of 150 mm focal length at f5.6 is 0.03 mm The difference in average of focussing of 150 mm at f16 and f5.6 was 0.065 mm

The focal length for f16 was 0.065 longer than at f5.6, but t-test showed it not to be significant because of the scatter of the data. Problems of film flatness are more important here.

Film flatness I have studied also in a fidelty hoder (4x5) and in a Sinar holder with pressure plate. The flatness was within 0.3 mm in the fidelity holder and within 0.1 mm in the Sinar holder. Both are within the 0.33 depth of focus at f16 of the 150mm lens at the extreme demand of a circle of confusion of 0.01 mm. (Normally a circle of confusion for 4x5 of 0.1 mm is used, which is too tolerant in the light of modern optics I think)

Hope this is of use.

[Stan. Laurenson-Batten]

My thanks to all contibutors for a very well conducted informative discussion on a oft delicate subject.

[grepmat]

The tests and commentary are interesting, but all I can say is... I want an 8x10!

Cheers.

[Phillip Liebowitz]

I don't understand why no one on here is pointing an obvious glaring problem with this study. I'll start by pointing out that the lense used on the Digital Rebel is an $80 consumer zoom that wouldn't compare to a $2500 Leica lense under ANY condition. This whole study is a complete farce: comparing $5000 film setups to $700 digital cameras is totally absurd. I hope that you all enjoyed this pointless exercise in self-justification which seems to be the hallmark of every high-end electronics purchaser discussion: I spent $10,000 on it, so it MUST be awesome. Give me a break.

[Michael Chmilar]

The issue of lens quality on the Digital Rebel was asked:



Ellis Vener: [...] with the dslr the problem is the lens [...] and [...] with the DSLR he uses a lens that isn't equal to the quality of the lenses used on the film camera,

Darin Cozine: The LF equipment was tested with top-of-the line optics. Meanwhile the digital cameras and lenses were entry-level.


and answered:



Bert Otten: Today I have looked with a microscope at the air image of the plastic lens of the Canon (18-55) at the same focal length setting as in the test and seen that on axis the image of the USAF chart is resolved with the finest detail blocks at full opening of the lens. This implies that the lens has a resolution of a factor of at least 2.5 times higher than the test result of the whole camera. The lens was not limiting.

[Phillip Liebowitz]

So according to Bert's microscope, there's no difference between the plastic zoom and a 50mm prime lense. I don't need a microscope to tell me that is totally absurd. ABSURD! A joke!

[Michael Chmilar]

I see no mention of a "50mm prime lens". Perhaps you should read the article?

[Phillip Liebowitz]

Michael,

Bert is claiming that the consumer zoom was not a limiting factor. By making that claim, he is implying that a better lense (such as a 50mm prime) would not improve the results of the digital test. I was simply stating the obvious about Bert's claim: THAT IS ABSURD! A JOKE!

Given the huge holes in Bert's method, it's clear that he set up this so-called study to give him the results he wanted from the beginning. It's a clear exercise in self-justification masquerading as an "objective" study. It's appalling. Sorry to rain on your little large-format circle-jerk, but you guys came looking for a study that justifies your big money purchases and Bert gave you what you wanted, so you just accept it without even questioning the obvious flaws in his "research."

[Nick_3536]

No he's claiming that the zoom lens is not the problem. That a better lens wouldn't improve the results. He's not talking about the quality of the zoom lens.

Big money purchases? What are you smoking? I've bought whole LF cameras and lenses for less then $80.

[Michael Chmilar]

In the discussion above, all of your points are openly discussed, in a less sensational manner. You haven't made any contribution.

Please point me to your "study", and subsequent discussion. I will be happy to read it.

[Phillip Liebowitz]

One of the film setups is a Leica M6 which, with the lense, would retail for over $4000. Good luck getting a Linhof for $80, buddy.

I think I've wasted enough of my time here. Have fun blowing your money so your photographs look good under a microscope!

Fools!

[Michael Chmilar]

Goodbye, Troll!

[Paul Butzi]

"So according to Bert's microscope, there's no difference between the plastic zoom and a 50mm prime lense. I don't need a microscope to tell me that is totally absurd. ABSURD! A joke!"

No, the statement is not that the plastic zoom and a 50mm (sic) prime lens are indistinguishable.

In fact, what he is stating is that ON THE DIGITAL CAMERA, the resolution of the plastic zoom lens is not the limiting factor - that is, the results from the on-axis tests with that camera would not shift appreciably even if the lens was improved dramatically.

It's easy to sit back and take potshots at Bert's results based on your pre-formed opinions and prejudices. It's another to produce useful remarks that are based on what he's actually done (for example, I don't see any 50mm prime lens used at all).

I think Bert is to be commended for having done such a thorough job. In particular, I appreciate the amount of work he's done to eliminate focussing as a variable - just the process of measuring the repeatability of focus, the focus shift, etc. is a considerable amount of effort and I thank him deeply for investing the effort to control this often ignored variable. There's a world of difference between an experimenter telling me that focus shifts are not significant, and an experimenter telling me he's used t tests to evaluate the difference between the focus means and found them to be not significant. Given that these tests were performed for the film based cameras, I wonder to what extent focus errors caused by alignment issues (and by inadequacy of the focusing system itself) have affected the results from the digital cameras. For instance, what is the focus system used by the Canon S50, and is the lens focused continuously or by zones, as is done with so many film based autofocus point and shoot cameras?

One thing I find missing is the lens to subject distances for the various setups - I assume that it was adjusted to get approximately the same framing after accounting for differences in aspect ratio?

Beyond the obvious conclusions about spatial resolution from Bert's tests, I'm finding it very interesting to evaluate the different images in terms of noise. For instance, although it's no match for provia 100-f in the Leica M6/35mm Summicron combo in terms of spatial resolution, the Eos300D seems to be turning in a pretty impressive performance in terms of noise, although it suffers quite visibly from what appear to be sharpening artifacts. However, Bert, you've offered no indication of whether you've used multi-sample scanning with the film - I've no idea if the Epson 4780 software offers a choice of multiple samples but I'm pretty sure the Minolta software does. After making some (probably unsupported) assumptions about the noise sources in scanning I would advance the proposition that the noise in scanning (across at least part of the tonal range) will fall as the square root of the number of samples - that is, quadrupling the number of samples taken will reduce the noise by roughly a factor of two. We could probably have a lively discussion on this, as well as on whether multiple sample scanning will improve or degrade spatial resolution.

I'd be particularly interested, Bert, in reading your thoughts on this as well as your thoughts on how much advantage of 8x10 is derived from more information on the film, and how much is derived from the fact that the larger negative simply eases the task of the scanner. It would be very interesting, for instance, to see 4x5 and 8x10 results drum scanned, to assess how much the scanning technology limits performance for 4x5 and 8x10. If you're interested, Bert, I will try to round up someone with a drum scanner and see if we might get drum scans of the 4x5 and 8x10 transparencies to add to this excellent article.

[Paul Butzi]

Naturally, no sooner do I hit the 'post' button that I immediately come up with more questions.

Bert, in the overall photo of the target scene, it looks to me like you've taken some care to have the plane of focus run through the beer bottle and through the USAF test target. Did you take steps to ensure that the film/sensor plane was parallel to this plane of focus, or did you do any analysis to ensure that the depth of field rendered this problem insignificant? Also, to what extent do you think the results have been affected by lack of field flatness for the various lenses? The 240mm and 150mm Rodenstock lenses presumably are well corrected and have very flat fields, as I expect the Summicron does. I have no idea how much field curvature there is with the 18-50mm zoom nor with whatever lens is in the S50.

In any case, though, I'd dearly love to see a similar test done for tonal resolution - perhaps including a reflection step wedge in the target scene with patches large enough to do RMS noise measurements at several different tonal values? A friend of mine has done a wonderful series of photographs with an Eos-1DS and my first impression of the prints is that the tonality from that camera just whips film all hollow. Now, before people get their knickers in a knot, I'll add the rather large caveat that that statement includes the word 'impression' and is not based on objective measurements or even direct comparison.

[Bryan Willman]

1. Darin Cozine reports "I have seen another test which compared a Canon EOS 1d 12mp with its counterpart EOS-1n. The same L-series lens was used, and various films were tested. The film was scanned with various methods and a microscope with a digital camera. "

Is that study on the web? Perhaps we could have a citation to it?

2. Bert Otten suggested that someone ought to repeat this test. I intend to, though a somewhat different one. (And it will be October before I get to it.) I propose to test 4x5, 35mm Leica, an Eos1V and an Eos1DS. The later two can share the same lens, be focused with similar autofocus or manually with the same 5x magnifier, etc. I don't happen to have an 8x10.

I also propose to try more scanners, possibly including an Imacon.

My (current) "experimental hypothosis" is:

A. Well scanned 4x5 will win. But getting a scanner to get all the data in the film will be hard.

B. Well scanned 35mm film will be better than the 1DS's 11mp image, but noiser, dirtier, and much harder to get. And the scanner capable of this result may be prohibitively expensive (twice the cost the already very costly 1DS)

bmw

[Kirk Gittings]

Thanks for a great test. It confirms my anecdotal experience. If you don't mind I will share this with my students next summer at the Art Institute of Chicago.

[QT Luong]

With regard to the criticism about using the cheap 18-55, using Norman Koren's excellent program (imatest.com), I have found that there is about a 10% on MTF 50 (measured on a Digital Rebel) between a 18-55 and a 24-70/2.8 when both are used at 24mm and f8. Many people have reported that the 24-70/2.8 compares well to a 50 prime from Canon. So the lens, even on a Digital rebel is a limitation, although a small one. As far as 6MP camera goes, tests and subjective reports have repeatedly shown that the Digital Rebel is on the top of the crop.

[QT Luong]

A couple of more comments: the reason why a better lens than the 18-55 changes the results even though the resolution of the 18-55 is already way above the resolution of the sensor is that the resolution of the captured image is a product (convolution) of the lens and sensor resolutions.

Many factors contribute to print quality, high-contrast B&W resolution being only one of them. So it is entirely possible that the 300D would actually produce a better print than the 35mm/5400 when used on ordinary photographic subjects.