Resolution limits of prints

[paulr]

I just wanted to post some observations that I've been making while scanning a mountain of negatives and prints for a book project.

It seems that the prints are all resolution limited by the paper surface itself. Beyond 300 or 350dpi, it seems impossible to extract any more detail from the prints. However, larger prints of the same image exhibit dramatically more detail. In either case, the resolution limit is not obvious with the naked eye, but it is evident with a loupe or with a scan that's zoomed in on in photoshop.

Example: I have an image that is printed at my normal size of 11.5 x 9, and also in mural size at 48 x 38 inches. The negative is 4x5, apx 100, very sharp. The mural was printed with an HK horizontal 8x10 enlarger with a modern 300mm rodenstock lens and a vacuum easel; the small print was printed normally with a schneider apo componon hm lens. These are both modern lenses in well aligned enlargers, and can be expected to offer comperable performance. If anything, the small lens should be better as it is half the focal length (made for the format being used) and an apo lens.

In the mural, there's a blackboard clearly visible inside a partially demolished building that has bowling scores written in chalk ... "individual records, 1932" and a bunch of numbers. On the small print, no matter what the resolution of the scan, or the power of a loupe, the writing is an illegible blob. Much beyond 300dpi, there's no additional revelation of detail, beyond those of thee paper surface itself. This corresponds to around 700 to 900 dpi of useable resolution on the negative (which might be defined as resolution that can be rendered at more than 20% or so contrast on an MTF chart). In the mural, however, the letters were so so crisply renered that they have edges on them. I measured cleanly resolved writing at 1.5lp/mm, which must have preserved at least 50% contrast. This translates to 38lp/mm at a respectable contrast, an equivalent of 3600 dpi of EASILY useable resolution on the negative. Which is about 4 times what was available on the print that was ... one fourth the size. If anything things look better for the big print, which has two possible advantages: smoother paper surface (very glossy RC paper, vs. air dried fiber for the small one), and a larger magnification which uses the lens closer to what it's optimized for (my apo componon is optimized for 10x enlargements, but I'm using at about 3x).

So? I think this is helpful information if you're scanning for any kind of printed project. Resolution beyond 300 or 400 dpi at final print size will not be visible, because the paper surface is limiting.

A related point: if paper were not a limiting factor, the next limit is human vision. I learned from a tech rep at Schneider that human eyes are limited to resolving 11 lp/mm at the best viewing distance for fine detail (12"). This has to do with the density of light-receiving rods and cones in our retinas, so it cannot be improved optically with glasses, etc.. This translates to close 558 dpi of resolution, or close to twice the paper limit, if your eyes are perfect. However, this is DETECTABLE resolution. Meaning we might be able to see 11 lp/mm details if they are rendered at close to 100% contrast, and yet we would probably perceive them at a ghostly 5% resolution (about the lowest contrast that's visible).

The Schneider rep told me that the important number is 5 lp/mm--this is the resolution range that our eyes use to perceive sharpness. Which means that having strong contrast at 5 lp/mm (around 250 dpi) at print size maters much more in our perception of clarity and sharpness than the rendering invisible or barely visible details at higher resolutions.

[Tim Curry]

This has been my argument about 4x5 enlargements vs. 8x10 contact prints. The resolution of the paper is what limits size and image quality in less than 10x enlargements (depends on actual film, grain, developer, etc.). At a certain point, the image will become soft if enlarged too much, but a 4x5 film enlargement to 8x10 is the same as an 8x10 contact print if the scale is the same. The film is not the issue, it is the paper which limits in this case.

[paulr]

Tim, I've thought about that a lot myself. I make a lot of 4x5 contact prints, a lot of 8x10 enlargements from 4x5 negs, and have looked at a lot of 8x10 contact prints. I have to conclude that the contact prints do look different to me. My post was about resolution being limited, but there are many other factors besides resolution that influence how sharp a print looks. I don't think the 8x10 contact will reveal any more detail than the enlargement, but it seems to have different qualities. Going back to the idea that perceived sharpness has a lot to do with contrast at 5 lp/mm, then it could be possible that the contact prints are producing more contrast in that range. The differences are subtle! Subtle enough that I never was tempted to get an 8x10. But I'm often aware of something being different .. maybe more real .. when looking at a good contact print.

[Neal Shields]

Ctien has done a lot of research on this and published his results in both Phototechniques magazine and his book "Post Exposure"

Beyond that there is much additional data in "Basic Photographic materials and Techniques" published by the Rochester Institute of Technology and also in the book image clarity.

All of these sources seem to show that photographic paper is not the limiting factor in print resolution and that 30 lp/mm is regularly obtainable by good practice. Furthar, studies show that viewers will precieve a photograph with a level of resolution equivlent to 30 lp/mm as "sharper" than one with say 15 lp/mm.

While the human eye is quite limited in telling a "dot" from a "point" it is highly acute in telling the difference between edges made from those dots and smaller dots which is why 30 lp/mm is precieved as sharper in side by side compairisons when it is impossiable for the human eye to tell the difference in the size of the dots.

Most comercial prints today are made on machines that print at 300 dpi so if a print is made on those machines from 35mm, another from medium format and another from 4x5 there will be minimal differences in the final print. The limit is in the laser writing and computer program not the paper.

[S. Preston Jones]

I know some photographers who put their 8x10 negatives in an 8x10 enlarger and make approximately same size prints and say that they cannot see any difference with 8x10 contact prints of the same negative. Do others out there do this and if so can you see any difference. I would think that the new optical system should add some differences or is this an incorrect assumption..

[Oren Grad]

Thanks to Paul for an interesting and stimulating post.

For anyone who's interested in this topic in the context of traditional photo materials, I'd strongly recommend the excellent article "Is Your Print Paper Sharp Enough" by Ctein, which appeared in the Mar/Apr 2002 issue of Photo Techniques. Ctein's key findings, based on some very careful experiments as well as long experience with many different printing materials for both color and B&W:

* Although 8-12 lp/mm is about the limit of detail that the eye can discern, the appearance of sharpness is actually affected by information up to about 30 lp/mm. That is, if you place a print that resolves 10 lp/mm next to another print of the same image that resolves 30 lp/mm, the 30 lp/mm print will look crisper than the 10 lp/mm print, even though the eye cannot discern any more distinct details.

* All currently-available print materials that Ctein tested, in both B&W (including fiber and RC, graded and VC) and color, resolve more than 50 lp/mm, often far more. A consequence is that paper resolution does not in any way limit visible print resolution or apparent sharpness. Interestingly, there was one material available in the not-too-distant past for which this was not true. Kodak's Ektaflex print materials of the 1980s were limited to a resolution of 18-22 lp/mm, and indeed were incapable of producing prints that looked "perfectly sharp" to the eye.

One thing that is clear is that at least in projection enlarging, it takes exceedingly careful technique - far more meticulous than most people imagine, and far beyond what any of us can manage on a day-to-day basis in real-world practice - to come close to achieving the maximum resolution and apparent sharpness of which our materials are inherently capable.

Specifically with respect to the comparison of enlargements from 4x5 with 8x10 contact prints, I am agnostic on the question of whether it could ever be possible to make a 2x enlargement from a 4x5 negative that is indistinguishable from a good 8x10 contact print. However, it is certainly true that in very careful experiments in my own darkroom, I have been unable to do so. For practical purposes, then, if I want the "look" of a contact print, the only way to get it that's available to me is to actually make a contact print.

One interesting consequence of the 10 lp/mm limit for visible detail is Paul's astute observation that you have to enlarge a negative in order for the viewer to see all the detail that was captured in it. But visible detail and apparent sharpness aren't the same thing. The paradox is that in general, you can't have maximum discernible subject detail and maximum apparent sharpness at the same time, because the enlargement process needed to pull all the subject information up to 10 lp/mm or less typically kills the information at 30 lp/mm that provides the cues for "perfect" sharpness. So we face an interesting choice - for perfect sharpness we generally need to make a contact print, but to convey maximum information to the viewer about spatial detail in the image, we need to enlarge. Which approach will work best for a given picture is up to the craftsman; different people will make different judgments.

Before anyone cranks up the old, pointless debate about whether contact prints are "better", I want to add that in my view, although the contact print "look" can be exquisite, isn't the only sort of look that can be very beautiful. There are other ways of achieving beauty than through the theoretical maximum of resolution or sharpness. Even with enlargements, we are fortunate indeed that the materials and methods that are available to us are sufficiently forgiving that we can make beautiful prints with a level of precision in techinique that, while it falls short of laboratory standards, is practical for us on a routine basis.

Digital techniques - including digital capture, scanning and inkjet printing - will force us to learn new rules of thumb for understanding departures from theoretical perfection, and their implications for our work. In particular, it seems pretty clear that with the combination of inkjet and matte- or art-surface papers, the print medium will indeed place some visible limits on resolution and apparent sharpness (but of course some pictures will work well in this mode anyway). I don't know anything about the behavior of glossy inkjet print media in this respect, though. My own experience with digital media is pretty close to nil, and I have a huge amount still to learn about the effect on image character of the various forms of sampling and quantization involved in digital imaging processes - image capture pixels, scanning pixels, inkjet dots, and so on.

[Oren Grad]

For Preston Jones -

I have tried experiments comparing a 1:1 projection print of a 4x5 negative with a contact print of the same negative. I am unable to make a 1:1 projection print that, to my eye, exactly matches the contact print. In my experiments, at least - with me, as observer and judge, being part of the experimental system - there is clearly some loss of apparent sharpness occurring through the projection process.

[paulr]

I'm very curious to see how the tests were done that suggest we can tell the difference between prints that resolve 15 lp/mm and 30 lp/mm in prints. I would bet that they were not controlled scientific experiments, since constructing such an experiment would be very, very difficult. It would require you to make pairs of prints that were identical in ALL respects besides maximum fine line resolution. It would be difficult to do this, and difficult to prove that you did without performing the same process on sinusoidal mtf charts and testing them--not a project within the grasp of a do-it-yourselfer likc Ctein. On the other hand, it's very easy to see how an alteration in printing process that would produce finer resolution could also affect contrast at the lower resolutions that strongly influence our perception of sharpness. For this kind of subject, which lies in the hazy overlap of physical science and human perceptions (like psychoacoustics) experiments are very, very difficult to devise and carry out. I don't pay much attention to results like these unless they appear in peer-reviewed journals, and unless the point being supported is specifically the one the experiment was designed to question.

As far as paper not being resolution limiting, this is a much simpler question, but I'm still curious to see how it was answered. If the tests were done with 100% contrast test targets, and the paper was considered to resolve the image if any hint of the bar pattern was detectable (which is how most resolution tests are performed) then I can imagin it being possible. But in a sense that'ss talking about the paper performing at 3% or 4% modulation, which is essentially useless. At these levels you're not depicting detail; you're making background noise.

My own tests are unscientific, but have the advantage of using real live images. I find it telling that there's no aparent maximum resolution difference between a contact print, a 1.5X enlargement, and a 2.5X enlargement with my negs. This is using a pair of eyes, a loupe or a professional scanner, and negatives made of subjects with real world contrast levels, not test grids. To really answer the question we'd need to see an MTF test ... but unless I'm missing a huge variable, the paper is acting as strong resolution limiter for real world images under real world viewing.

[Neal Shields]

I don't have it in front of me but there are very precice ways of statistically validating jury testing when individules are makeing compairisons.

There is a whole chapter on this in "Basic photographic materials and processes".

It basically deals with giving people choices that can be contridictary. I.E. if A is better than B, and B is better than C, then A must be better than C if not then your data isn't valid.

In my last post I forget to mention the book "Image Clarity" which is also a good reference on this subject.

Most of this stuff has been researched dillegently for years and different researchers with different goals in different places have confirmed each others results.

For example Katz's formula for clumulative losses through optical systems goes back about 100 years. It may not "prove" that a contact print will be better than an enlargement, but people have had plenty of time to disprove it and haven't yet.

[paulr]

"here is a whole chapter on this in "Basic photographic materials and processes".

It basically deals with giving people choices that can be contridictary. I.E. if A is better than B, and B is better than C, then A must be better than C if not then your data isn't valid."

Sure, this proves that there are perceptible differences between the prints in question, but it doesn't prove what factors led to the differences. It doesn't tell you if your higher resolution print isn't different in some other aspect as well.

I have the image clarity book, and have followed a lot of the discussions and articles on the topic. The only information that I've found that seems to be based on rigorous science is from the lens companies themselves. They need to choose appropriate compromises between resolution and contrast for different lens designs and magnifications, and you can be sure they've done the research on what's going to look sharpest before investing their millions. A lot of this information is unpublished, but ocasionally you can get your hands on some, especially if you make a friend at schneider or rodenstock ...

[Oren Grad]

Paul -

Ctein's paper-resolution experiments were conducted by contact-printing a glass bar target, using sufficient pressure - "equivalent to contact-printing an 8x10 negative under several hundred pounds of pressure" - to insure good contact with the paper, and exposure tuned for optimum resolution. He describes the bar target as having "nearly 100% contrast out to about 100 lp/mm; thereafter, contrast falls off until it reaches zero at about 500 lp/mm". The resulting prints were examined under a microscope to determine the resolution limits of the papers. He doesn't specify in the article the exact standard used to visually judge extinction of the bar pattern. However, I doubt very much whether he's fallen into any methodological traps here - Ctein is a first-rate amateur scientist and understands well the very legitimate issues you raise, with respect to both experimental design in general and psychophysics in particular. And his conclusion is that paper resolutions are *far* beyond the point where it makes a difference visually, a finding that should be robust to reasonable variation in judgment thresholds among experienced observers.

[paulr]

Thanks Oren, that's very helpful. It also confirms what I suspected, and is a near perfect demonstration of why resolution data, by itself, is useless ... unless you're buying a spy satelite.

I believe the results of his test completely, and for reasons that I stated in my post earlier, believe that those results are essentially useless and have little bearing on real world photography.

I'll go over the reasons again:

1) The test target, like all resolution targets, is a hard-edged, 100% contrast subject. Very little of what you photograph in real life, unless you operate a process camera (if they exist anymore) is anywhere near 100% contrast. And few edges in nature are as abrupt as test target edges.

2) The results were evaluated by microscope, and while he doesn't seem to mention how resolution was determined, I'm willing to bet that it was based on line pairs being distinguishable from each other in any way. Under ideal viewing conditions of a test target, this could mean that lines rendered at 1% or 2% contrast were considered rendered. In real life, however, you would not consider those lines to be rendered detail; you'd consider them to be hazy background noise. Under normal viewing conditions, fine detail is invisible below 5% contrast; it doesn't look like real detail until around 20% contrast; and it doesn't look sharp until around 60% contrast.

If the MTF curves of the paper were to show a long slope, declining below 5% contrast after the resolution exceeded 150 lp/mm, it's likely that you would see no detail finer than 150lp/mm in a magnification of real printed negative--even if test target detail might still be detectable under ideal conditions at four times that resolution.

3) This part I didn't guess--his use of a press to force the test target onto the paper may well reduce the most significant resolution limiting characteristic of the paper: its texture. If a physical press flattens the surface of the emulsion and guarantees near perfect contact, it would make sense that you'd get higher quality results than you might with more conventional (real world) techniques.

[Oren Grad]

Paul -

Thanks for your response, I appreciate the prod to be more careful about defining the relevant issues.

First, I agree, to really understand what's at stake here we should be talking about MTFs rather than resolution, which even when unambiguously defined with respect to contrast is only a single slice through the overall modulation transfer characteristic.

So more precisely, the fundamental question is, are paper MTFs limiting factors with respect to any important perceptual characteristic (discernible spatial detail, degree of apparent sharpness, anything else?) of pictorial images made on the papers?

At what spatial frequency does a paper's MTF decline below the (frequency-dependent) threshold of perceptual relevance? I speculate that Ctein's stated resolution figures, even if they do represent minimal contrast (which is a speculation on your part, and which is not certain) represent a point so far out on the spatial frequency scale that they could be consistent with MTFs that are indeed perfect with respect to anything that matters visually. But I take your point that Ctein's experimental results, at least as reported, do not prove this, and do not directly resolve (no pun intended <g >) the basic question.

[paulr]

since MTF testing is beyond anyone's means at home, why not do an informal test with your own prints? Next time you have an oportunity to scan a print that you've made in different sizes (ideally one of which is a contact print) scan all of them at a very high resolution. Res them down in increments, and notice at what point you see detail becoming obscured.

There's probably a better way to do this test; if you can think of one, let me know. Maybe it would make sense to do multiple scans and different resolutions.

My observations came from scanning many prints, all different sizes, including contact prints. Many of them extremely sharp and detailed. I noticed that in all cases there was no difference in visible detail between 360 dpi and 720 dpi ... not counting paper surface detail. But large prints from the same negs revealed detail equal to many times this resolution. This was done on an epson 1640 XL (at work, sad to say).

A friend of mine who does copy work as part of his job and who's been doing piezography since the beginning has had the same findings. He works with medium and large formats, including 5x7 (enlarged) and 8x10 (contact) so I trust he knows what detail looks like.

Again, none of this is science, but it reflects a pretty systematic survey of real live photographs. Scientific experiments, which have to balance accuracy and repeatability with limited means, often end up designed in ways that fail to mimic the real world situations they're attempting to investigate. The resolution test in its classic form is one typical example.

[Oren Grad]

>> Scientific experiments, which have to balance accuracy and repeatability with limited means, often end up designed in ways that fail to mimic the real world situations they're attempting to investigate. <<

This is true, alas.

This interesting experiment will have to wait - I have no scanner at the moment, though I'm sure I'll end up with one eventually.

I'm a little nervous about the possibility of artifacts arising from the particular algorithm used to "res down in increments" from a high resolution scan, and also about artifacts that may arise when sampling frequencies start to approach critical spatial frequencies in the image. Doing original scans at different resolutions, as you suggest, may shed some light on this, although one would still need to understand what the scanner itself is doing differently in hardware and/or software to achieve the different resolutions.

>> the most significant resolution limiting characteristic of the paper: its texture <<

This may be relevant to an observation I have made in my own experiments, that contact prints on glossy RC do seem to retain more information than comparable prints on glossy FB. I'm not sure whether this is a perceptual illusion or an actual difference in fidelity of reproduction under my contact printing setup, which provides pressure that's adequate for my purposes but not anywhere near what Ctein applied to assure optimal contact in his experiments. I do have a high-quality resolution test slide floating around somewhere; it might be interesting to dig it out and try contact printing it to see whether there's an actual loss of information in an FB compared to an RC print under my usual working methods.

[Henry Ambrose]

Assuming that you are scanning your own work and that it does not involve 37 steps and weeks in the darkroom making prints, I wonder why you are not scanning your negatives? Removing the optical printing generation will make your job lots better. You could completely avoid the limitations you describe in scanning prints and end up with very high resolution files with more detail than you can put on paper in a book.

What process will you be using to reproduce your photos for your book? The real task is to determine how much resolution you need to reproduce all the detail that your ink and paper will display and of course the size of each image in reproduction. Now if you're exercising your mind, wondering aloud or you are working with historical material and have only the prints, pardon me for butting in. Otherwise, scans from negatives and press proofs are in order to determine what works for your print job. You really do need to start with the desired end result in mind and work backwards from there. - which seems to be left out of the conversation until now.

As to settling the issue of print resolution with strict testing and MTF graphs, I'm pretty sure that if you look at some photos you can pick the best example. Its pictures - and the one that looks best - is best.

Finally, from the original post: "but unless I'm missing a huge variable, the paper is acting as strong resolution limiter for real world images under real world viewing." The paper is not limiting resolution for viewing by human eyes, but from lots of experience I can say that paper prints will limit resolution for further enlargement. And why wouldn't it? You're making a copy of a less than perfect copy.

[paulr]

Hi Henry, thanks for your post. I definitely agree, especially with your conclusion that the eyes are (for all practical purposes) the limiting factor when it comes to fine detail resolution in prints.

I wrote the original post not to describe a problem I'm having, but rather to let people know that there's probably no reason to scan prints at computer-choking resolutions (or to scan negs at a resolution that will be huge at their final print size).

And also because I thought it was interesting.

I'm actually going to do a combination of scanning negs and scanning prints for my project. I'll scan the negative that haven't been printed already, and also the negs of images that have too much important detail in deep shadow--the scanner doesn't seem to do a good job of seeing that far into shadows, even in 16 bit mode.

Otherwise, I'll scan the prints, since the results are more than sharp enough, and because I've already done hard time making the prints and getting the values where I want them to be.

The book is going to be a small edition artist book, printed ink jet with piezography. The book prints will be the exact size of the original prints. I may end up redoing scans if I'm not getting the results I'd hoped for. In some cases the scans from the prints may not translate well into the tonal scale of piezography, in which case I'll have to go back to the neg and rethink the whole print.

I've been scanning at 400 dpi at final print size, which is a little higher than the highest resolution that seems to make a difference. I like to leave a little headroom to minimize the effects of any future manipulations. All the scans are 16 bit.

Scanning the negs has so far been easy, but I've only made fpo scans at a relatively low resolution. I'm sensing that I may not know what I'm doing as far as maing the best final scans from the negs.. If there are any tricks to getting excellent scans from bw negs (in terms of getting the tonal scale accurate, not in terms of resoultion) ... or any good books I should read, I'm open to suggestions.

[Henry Ambrose]

Paul, You are taking on a job that used to require several highly skilled people and some expensive equipment! At least you don't have to worry about color separations since its B&W.

Its been a year since I've made any Piezo prints. My printer head clogged for the last time and I said "to hell with it" and have not touched it since. When it worked, I did make nice prints, some in a book of prints documenting the construction of a library that was well received. (about 27 images, edition of two, one for the donor and one for the institution) I found with my printer set up (the old Piezo plug-in for Photoshop and a 1200 with Piezo inks) that more resolution up to about 720 PPI at repro size seemed to show in the prints.

When I tried to scan darkroom prints I found that intricate detail got lost unless I printed about half size of the original - then they looked pretty nice but still not like a good scan of a negative. The reason I tried this was to see what would happen when attempting this workflow, knowing that it was likely not a good solution. I was right about that part! I've scanned many prints, negatives and produced thousands and thousands of digital captures for offset reproduction and done some historic restoration of old prints so I had a good idea of what to expect. I was fascinated at the time with the observation that inkjet printers could often make a very good looking print from what I considered to be poor quality files from digital cameras. You have probably observed this - a file with all kinds of artifacts, jaggies, softness and flaws that makes a good looking Epson print.

If you are working with negatives that printed easily in the darkroom, they will probably scan nicely as well. I don't think any book will help that much with scanning. (there may be one but I don't know of it) Not nearly so much as just doing it and trying different settings and adjustments. I usually scan in 16 bit after setting black and white points and bring the file into Photoshop to do most adjustments starting with global levels and curves then spotting and going on from there. Depending on your scanner and its software this approach may work well for you - only course adjustments in the driver then manipulate as desired in Photoshop at 16 bit. A very nice side benefit of this method is that you free yourself from the vagaries of different scanner software and bring your workflow into one unified stream. This helps a lot if you use more than one scanner or you have one that is supplied with crappy software.

Good luck on your project!

[paulr]

Thanks Henry, you're right that I have my work cut out for me. Soon enough I'll get to compare scans from prints with scans from negs ... I'll post the results. In the mean time I have other worries, like editing and sequencing ten years of work, learning bookbinding, and getting someone to pay for all this!