Published 2010/10

Introduction

This section is tantamount to "pixel peeping" but it seemed there might be some value in exploring this subject with a view to estimating what can be accomplished with various types of digital cameras and film. Avoid reading if you just want to get on with taking photos.

For a somewhat different approach, arriving at much the same conclusions, check Thom Hogan at this link or Norman Koren for a really excellent treatment of the subject with all the technical details you could want.

The numbers and calculations you will find here are approximate and/or rounded. They are by no means critical in any event. There's considerable latitude in how all of this is interpreted by the eye. At a comfortable viewing distance even a "poor" resolution print might prove enjoyable. Resolution isn't everything and it's more important in some images than others.

Film and Paper Resolution

In digital photography the image is actually an array of millions of dots (pixels) each corresponding to an individual light detecting element on the camera sensor. The image might later be re-sampled in software so that the one-to-one correspondence no longer exists but it will still be a mass of dots. Unless a digital image is excessively enlarged, the eye (and possibly the presentation medium - paper, especially) blends the dots together and the image appears continuous, without any visible underlying structure. The number and distribution of pixels sets a strict limit on the amount of detail a digital image can represent. If the pixels are too far apart, it is evident some detail will slip through the cracks and if they are too large, detail that is finer than the size of the pixels will be blended and lost.

Film is an analog medium and doesn't have pixels or discrete light sensing elements. It is seemingly capable of recording an infinite amount of detail. Such is not the case, however. Among other things, light spreads out a bit upon entering the film, spilling into areas adjacent to where it first struck, compromising detail. Grain is always present to some degree and this, too, limits just how much detail can be recorded. Finally, all lenses, even "perfect" ones, are significantly limited in their resolution by their finite size apertures and other factors.

Years ago, working with test patterns, I determined that a 35mm color film frame taken with good equipment had an equivalent 24 or 25 megapixels capacity (you will find this number confirmed in technical literature on the Web). A 35 mm frame is 24 X 35 mm so  24 megapixels translates into a linear pixel density of about 170/mm (85 line pairs / mm). There is considerable variation in this number depending on the film type but it's in good agreement with Koren's findings for the sharper color films.

Note that for estimating resolving ability, the total megapixels is a completely useless number by itself. It is the "pixels/inch" or "pixels/mm" that counts. You never see that number in camera ads - probably because it is a small number and small numbers don't impress people. For meaningful comparisons you have to know (or figure out) the linear pixel density. Digital Photography Review (dpreview.com) now publishes a pixel density (pixels/square centimeter) number with each camera review. This makes sense - at last. To get the even more useful pixels/mm number, take the square root of dpreview's number and divide by 10.

There are specialized films and lenses used for copy work, microfilming and so forth that have a lot more resolution than this but they aren't suitable for general photography and won't be discussed here. We'll just assume that something like 170 pixels/linear millimeter is a reasonable effective pixel density estimate for film working at or near its best in good equipment under realistic conditions. 

So to continue ... from this linear pixel density we can calculate that ordinary film (of any size) can provide the equivalent of around 4300 pixels/inch. Now, recall from elementary signal processing theory that for any sampling rate, the maximum resolvable detail is half as much. In other words, 4300 pixels/inch can just resolve 2150 "line pairs"/inch (lp/i), a line pair being a black line sharing a boundary with a white line. Now this gets interesting. Suppose you want to make an 11 X 14 print. You have to enlarge that 35 mm frame by a factor of 10. You now have a maximum 215 lp/i resolved because the pixels have been enlarged too. Papers have highly varying degrees of resolution but an image will look about "as sharp as it gets" to the unaided eye if the presentation is at 200 lp/i (8 lp/mm) on glossy paper. If you read Hogan you will find that ink jet printing limits you to something a bit less than this anyway. In my own tests, using mathematically generated line pairs, I discovered 200 lp/i about the limit as well, using a high quality ink jet and premium glossy paper.   There's a fair amount of latitude involved with all of this. Hogan uses 150 lp/i (or anything over that) as his standard  for "excellent".  That's equivalent to 6 lp/mm so any number from 6  lp/mm upwards in the final, enlarged image represents the potential for "excellent" ("outstanding" in Koren's scheme of things) print quality.  Someone who uses a magnifier will probably disagree with these numbers but here we are considering only comfortable, unaided viewing of images.

There are research articles on the Web suggesting that the eyes of younger people can reliably identify images rendered at 300 lp/i (600 dots/inch) but the significance of this in the photographic context, where we are not usually submitting ourselves to a test of visual acuity, is probably small.

Computer Monitors and TV

Typical computer monitors and HDTV displays have a resolution about 1/3 that of inkjet printed glossy paper or 2 lp/mm at desktop size ("poor"). That's why it's silly to obsess over pixel counts if your intended presentation medium is a computer or TV. There's really no point in going into matters of resolution with these types of display. Why, then, do monitor images often look so great? Maybe Rockwell is right and high detail isn't as important to a pleasing result as we might want to think. It also helps that monitors are self-illuminating which provides greater control over contrast, saturation and brightness of an image than is possible with the natural, incident lighting by which prints are typically viewed.

35 mm Film

The numbers so far explain why it has always been something of a challenge to get a really detailed 11 x 14 from 35 mm film and, in fact, that pretty much represents the upper limit for fine prints made from an uncropped 35mm frame (for cropped images the situation is obviously worse). At 10 X enlargement, 35 mm resolution just about matches what the eye can perceive in a printed image. Even then, you have to get everything just right: perfect focus, a stable platform (or high shutter speed), a great lens and film capable of doing it justice. Ten times is a rather extreme magnification and the slightest defects are apt to become all too visible.

Of course, technique, the very best equipment and other variables can swing the results a bit to squeeze out some extra resolution by pushing film to its limit and I have seen satisfying 16 X 20s from 35 mm film - but very, very rarely.

Medium Format Film

Now imagine you are using a medium format film camera that produces 60 X 70 mm frames and that you are comparing results to 35 mm, using lenses of exactly the same focal length in each case. The pixel density per linear inch is just the same as for 35 mm even though the total pixel count is enormously greater. Film is film, after all, and how large a piece of it you use has nothing to do with how much detail it can record. The big advantage becomes apparent when enlarging. That uncropped 60 X 70 image enlarged 10 times will give a print 27 inches on the long side (but with a wider field of view) with just as much resolution as the 11 X 14 print (225 lp/i) from an uncropped 35 mm frame. Enlarged only 5 times to make an 11 X 14 the detail will be double what 35 mm can deliver but you probably won't see it because your eye (and probably the paper/printer combination) is incapable of rendering it.

Cropped Images

Let's consider cropped or scaled images. Suppose we have a 35 mm camera and a medium format, each working at the same focal length with the same type of film. Obviously, image magnification (ratio of image size to subject size) will be identical. That is, the image of the subject projected on the film/sensor is the same size for both cameras because they are using identical focal lengths. (The medium format image will be much wider-angled simply because the frame is wider.) Suppose further we want to crop out a central portion of the medium format image so we have exactly the same view vertically and horizontally as provided by the uncropped 35 mm image. It should be fairly obvious the enlarged print images will have identical resolution. The large format advantage is gone. The whole point is that frame size and pixel count each taken by themselves tell you nothing when it comes to resolution of the final presentation. A large frame has an advantage over a small frame (assuming similar resolution) only if you do not crop the former down to the size of the latter. In this example imagine that instead of cropping the medium format image to match the 35 mm we matched the 35 mm to the medium format by using a wider angle lens. If we enlarge to 11 X 14, detail should appear almost identical because we are essentially limited by the presentation method and the eye. But 35 mm has just about run out of gas whereas that medium format uncropped image could go to twice the linear presentation size with the same resolution.

Lenses

Nothing much is said here about lens performance but that's another crucial factor influencing resolution. If you check dpreview and photozone you will find many lenses  do not push the limits of film or, indeed the limits of a high resolution sensor such as that found in the D800/D7000. Excellent/Superb optics typically cost a fortune but will exploit what the best films and sensors can provide..

Digital Image Resolution

All of this has implications for digital. There has been quite a bit of discussion concerning "full frame" (meaning 35 mm film dimensioned) DSLR sensors as opposed to APS size sensors which are about 2/3 full frame. Until recently, the only true advantage of digital full frame (or FX, as Nikon calls it) has been lower noise because the manufacturers use larger pixels in these. That makes these cameras very attractive to professionals having to get action shots in available light. High ISOs to 6400 or more are entirely usable. Large pixels also deliver better dynamic range - an advantage in high contrast lighting.

The Significance of Pixel Density

A seldom mentioned fact is that, until recently, these full frame sensors actually have lower resolution than most smaller sensors for the same image size. That's because smaller sensors typically use smaller pixels, leading to a higher linear pixel density. Providing those smaller sensors have acceptably low noise (it's not going to be as low as on a full-frame/large pixel unit), the small sensor wins the resolution contest for identical image sizes and similar pixel counts.

On his site Thom Hogan points out that the bird photographer or others needing to significantly magnify/crop their images might want to think twice before using a full frame low pixel density digital camera. Photographs of that sort often require considerable enlargement of a portion thereof and linear pixel density is vital to preserving detail. In other words, at current state of the art, full frame digital (at less than 24 megapixels) may not always be the best choice if the subject of interest occupies less than the full frame and requires considerable enlargement.

Lately the game has changed. Sony's A900 DSLR (like Nikon's new D3X) is full frame and has high pixel density at 24 megapixels or 170/mm. There is now also the Nikon D800 FX with 36 megapixels and a linear pixel density of over 200/mm. This is doubtless a preview of the future for small format digital photography. It's very expensive technology but that should change in time.

With the possible exception of Sony's A900 and the Nikon D800, full frame digis currently do not match the theoretical resolution of film. Consider now a digital camera like the Nikon D300. It uses a smaller APS size sensor (about 2/3 full frame) with 12 million pixels. Do the math and you find the linear pixel density actually does match or even exceed what we find with film - at about 180/mm. What that means is that for the same image size (meaning the field of view on the APS frame will not be as wide),  taken at the same focal length in each case, the D300 should be as good as film when comparing prints with subject material at the same scale (for reasons not entirely understood, it may prove visibly better). The smaller frame size does mean, however, that theoretical resolution becomes a bit lower than that of film if you enlarge the full APS frame to the same size as an enlargement from an uncropped 35 mm frame. To make an 11 X 14 from a full D300 frame you must enlarge by 15 X resulting in 6 lp/mm resolved (still "excellent").

Compact Digitals

What about so many pocket digis now boasting 12 megapixels or more? These barely merit any discussion at all. They have among the tiniest pixels of all digis and should deliver fantastic resolution but this is masked by noise and noise processing within the camera. Furthermore, those dinky little lenses simply cannot provide the detail to begin with. It's all just marketing hype. Many digi compacts can, however, do better than cheap compact 35 mm film units - the competition isn't that tough. They don't need (nor can they benefit from) more than 5 to 6 megapixels to accomplish that.

Super-zooms

Like snapshooters, superzooms have small sensors and tiny pixels. The pixel size superficially suggests phenomenal ability to record fine detail. For example, the Panasonic FZ50 has a pixel density of over 450/mm. At the same image size this is almost 3 times the resolution of a full frame 24 megapixel DSLR (also 3 times the resolution of film!) but ... the frame is so small at just over 7 mm on the long side that it has to be enlarged 5 times to equal full frame size and that knocks the effective resolution (at full frame size) back to about 100/mm or 2540 line pairs/inch. You can visualize now that making an 11 X 14 print ends you up with about 130 line pairs per inch which, as noted earlier, is a bit less than what you can see in a glossy paper. A top-notch super-zoom like the Panasonic FZ30/50 - which has excellent optics able to put very fine detail on those miniscule pixels - can make very good 11 X 14's provided there is not too much noise (noise being an all too frequent problem, unfortunately) but that's about it. So, although a lot of detail can be recorded by the sensor the ultimate presentation size is limited by the enormous enlargement required. The principle imaging advantage of a super zoom over a compact is large, high quality optics.

Perceived Sharpness 

In my experience, DSLR and even some super-zoom images often surpass 35 mm film format in apparent sharpness at any print size. I can almost always easily make excellent large (11 X 14) to very large (13 X 19) prints from D300 and D80 images whereas using even "pro" 35 mm film equipment 11 X 14's are more apt to prove a struggle (whether in the wet darkroom or using scanned film). How can this be? This may not be the answer but it all probably has to do with the fact a solid-state imager is nothing like the "chemical" imager inside a film camera. It is possible that light scatters more within a film emulsion than within a solid-state sensor or that sensitivity to invisible wavelengths of light has a negative influence on film imaging. The full potential of film is, for whatever reason, more readily demonstrated in the lab than in the field. Digital images seem to possess an additional clarity and depth which may enhance the impression of sharpness. This difference in favor of digital becomes especially apparent in dull or otherwise poor lighting.

In general, whether evaluating film or pure digital images, perceived sharpness has to do with more than just objective measures of resolution. Among other things, contrast, saturation, brightness and, perhaps, even choice of subject play a role.

The Medium Format Film Advantage

Neither digital nor 35 mm film can support the production of satisfying huge images (beyond 13 x 19 or thereabouts) but to that limit, good digital equipment easily holds its own  For largest presentations, medium or large format film is still the only way to go and it's going to stay that way for a long time. It really is no contest. If you have never worked with a good medium format image (scanned or in a darkroom enlarger) you have no idea just how good film can be. It's astounding. There are a few "pro" digital cameras by Hasselblad and others offering medium format sensors with 50 megapixels and these purport to challenge medium/large format film. They cost tens of thousands and are beyond the realm of discussion here. Recently (late 2009) the Nikon D7000 has arrived and has the highest pixel density of any DSLR. The resolution of this camera encroaches on medium format film territory even for print sizes somewhat beyond 13 X 19. Yet more recently (2012) the Nikon D800 has gone well beyond what the D7000 can offer and it is safe to say that medium format imaging supremacy is now facing serious challenges.

This table summarizes all of the foregoing:

The terms "excellent"/"good"/"poor" are more or less in agreement with Hogan's definitions of print quality. If you visit Norman Koren's site you will find the standard somewhat more relaxed, with "Outstanding" added to the quality rankings. This only goes to show that even amongst acknowledged experts there is a some variation in judging image quality. Consider any findings regarding this subject as guidelines only.

Things to note:

• When enlarging the full frame from each camera to 11 x 14 size all the DSLRs deliver excellent resolution (i.e., as much or more than what the unaided eye can see). The D300 can still be expected to deliver "excellent" (if just barely) at the next larger standard print size but you might see the high density full frame units pulling away from it there.

• The FZ30 estimate is confirmed in practice. Very nice 11 X 14s are routinely possible but upon close inspection a touch of softness is seen creeping in, as compared to the D300 (for example). Its biggest problem is sensor noise which nibbles away some detail even at the base ISO. It's greatest asset is the Leitz lens which puts those tiny pixels to good use, placing the FZ30 well ahead of any compact unit although its sensor is no larger. As with compacts, dynamic range is rather severely limited.

• The whopping pixel density of the non-DSLRs implies a superior image but as we know, the tiny pixels are noisy and the small sensor means a correspondingly high enlargement is needed to get to presentation size. In the end, we find this pixel density leading to the sort of quality (at best) readily exceeded by the DSLRs.

• The Sony DSC-V1's resolution is suitable for fine 8 X 10's but 11 X 14 is largely out of reach. A couple of surprisingly "good" 11 X 14s have been made from this snapshooter - under ideal and rare circumstances. Despite its 2003 vintage, the low noise and fine optics make this unit something of standard, even today, for image quality in a compact digital. Its performance renders highly questionable the merits of the current "pixel race" in small digital cameras.

• Suppose we now shoot a subject whose image just fills the frame on the D300 (let's say it is a bird). We then shoot the same subject at exactly the same focal length with the D3X, D3 and A900. It won't fill the frame on those units but only about 2/3 of it although in all cases, the subject occupies precisely the same area on the sensor. We then enlarge the image of the subject from each camera so that it fills an 11 X 14. The subject will be exactly the same size on each of the three 11 X 14's. The result is in the last column of the table. The D300 wins the theoretical resolution contest in this case because of the higher pixel density but insignificantly so over the D3X/A900. The D3, however, falls significantly behind to "good". This is what Hogan is getting at by pointing out the value of pixel density for certain types of photography (small subjects, in particular). The super-zoom and snapshooter aren't included because their sensors are only a fraction of the APS frame and the subject itself would be heavily cropped - truncated within the camera itself. Because no cropping of the frame would be involved, their performance would be unchanged but you would see only part of the subject on the prints.

• The compromise APS format (D300) is a high value performer offering low noise, a reasonable frame size and high pixel density. It overcomes most of the problems with smaller sensors (noise and limited dynamic range) while being inexpensive relative to full frame.

For the ultimate in image quality, especially at very large presentation sizes, big film formats (provided you compose to fill most of the frame and don't crop too much) hold the high ground by a wide margin over both digital and 35 mm formats. Even at smaller presentation sizes they can prove much easier to edit and manipulate because so little enlargement is required. The high magnifications required by smaller formats - both digital and film - readily pose challenges in dealing with imperfections (dust, noise/grain, optical faults) because, of course, the flaws become enlarged along with the image and are more likely to be visible in the final presentation if not carefully managed.