Exploring the limits of 35mm BW Photography
I have been testing the full range of high resolution films since a number of months and what began as a simple filmtest, has now been expanded into a fresh exploration of film and lens capabilities. My results surprise me as much as I hope it will surprise the readers of this report. An open mind is necessary, however, as many established ideas have to be discarded! The Gigabit film has been promoted as a new dimenson in Hi-res BW photography and a useable resolution of 600 to 900 lp/mm (every line pair a black and a white line or space) has been quoted. In several usergroups this claim has been discussed to the extreme: could it be possible theoretically and if so could there be lenses that can use this capability. First the basic fact: the resolution of the eye: under ideal circumstances the eye can resolve at a distance of 25cm at most 6 to 10 lp/mm and here we take the absolute limit, which is reached when the eye is using its vernier acuity resolving power. Six lp/mm would be, according to all ophthalmic handbooks, a good average. This means that any detail which is smaller than 6 lp/mm cannot be detected as separate lines and will be seen as a grey patch. A simple calculation shows that to be able to see details that on a film are recorded with 600 lp/mm, we need a magnification factor of 100 times.
An absolute limit: the resolving power of the eye.
Here we have a limit that can not be brushed aside by whatever argument we want to use. Like the speed of light, this is a constant that limits in the end the demands we are able to define for our material and the specifications of the material, like films and lenses. There is much valuable discussion at the several Minox sites, where the maximum of resolution and fineness of grain is imperative to get the best results from the tiny negatives. The results are often amazingly good and one wonders how that is possible. Let me introduce a not so well-known phenomenon in optical lore: the best of Minox (sub miniature) lenses deliver a higher performance than the best of the 35mm (miniature) camera lenses. It is a physical fact that lenses that are quite small, can be corrected to a higher degree than their more voluminous companions. So the subminiature Minox lenses are closer to the heavenly state of being diffraction limited than others and with careful technique can deliver (relative) performance on print that will be the envy of the 35mm worker. "Relative" does refer to the final enlargement of the negative. When enlarging to a factor of 20, the Minox negative should bring an advantage, but when enlarging to the same print size, the 35mm neg will win.
These relationships hold of course too when discussing 35mm versus 120 (medium) format.
On the assumption that a 15 times enlargement is the limit for really critical work for the 35mm worker, even when doing gallery-fine art imagery, we can calculate the limit of resolution as being 15 times the limiting resolution of the eye. As this limit should allow for several types of eye-power, we should ue a bandwidth figure. The true limiting value is 1 minute of arc, as here we jump to the spacing and size of the cones and rods in the retina. The distance between cones is measured as 0.5 minute of arc and we need one additional cone to separate the detail and therefore the 1 minute of arc is the limiting value.
This has been established as scientific fact, but even here there are some who decline the facts.
Now one minute of arc at a visual distance of 25cm translates into 8 line pairs per mm. We simply cannot see finer detail at that distance due to the retinal structure. Let us however argue with 10 lp/mm (easier to calculate!).
At 15 times enlargement, we need to have on the negative a maximum of 150 lp/mm. And if we wnat to have some reserve capacity a figure of 200 lp/mm is really not only sufficient, but the maximum.
What about the resolution figures for document films?
Document films like Kodak Technical Pan, Agfa Copex Rapid AHU (AKA Gigabit in Germany), Maco Orthopan 25 and several others, do indeed have resolution figures of 350 to 600 lp/mm according to the manufacturer's datasheet.
There are two very important caveats here: this resolution is measured on negatives that have been developed to the extremely steep characteristic curve needed for halftone (black-white) reproduction in special developer chemicals.
the resolution figures have been produced by contact printing a glass plate with finely etched black/white lines with the emulsion. For very fine structures an even finer pattern can be used, again on a glass plate, with a chromium layer in which the patterns are etched.
Now to give some scientifically established data, I can refer to some publications that have made available their testing methodology in detail. The resolution figures for Technical Pan, APX25 and TMax100 and Panatomix-X have been published, based on the (ideal) contact print with glass plate and with developers for continious tone reproduction:
TP = 250 lp/mm, APX25 = 179 lp/mm and Tmax100 = 110/130 lp/mm, Panatomic-X = 180 lp/mm.
Not all methods are directly comparable, but we may safely state that under scientific testing conditions, an ideal limit of about 200 lp/mm is feasible.
The publications referred to, all warn that these figures are ideal, that is without taking into effect the imaging degrading effects of the photographic imaging chain.
There are several magazine reports and data sheets from a manufacturer that state without any reference to the method used to establish the published results, that even in continuous tone reproduction with normal photographic lenses, resolution values above 500 lp/mm can be usefully employed. These claims beg to be substantiated and without factual evidence should be approached with some healthy suspicion.
Resolution and contrast in practical photography: diffraction limit.
While it sounds very impressive to discuss resolution figures above 500 lp/mm, we may read the discussion by Zeiss:
They state that the resolution of very fine detail is limited by optical aberrations and the limiting value of diffraction. A lens that is diffraction limited is so well corrected in its optical aberrations, that only the diffraction of light does influence the quality of the image. The resolution of a diffraction limited lens (which is hardly encountered in photographic practice) depends on wavelength and maximum aperture.
Again using ideal values for the Rayleigh limit, and applying the contrast theory of Kühler, the resolution for the average wavelength of light (0,555 micron) is
Again these figures are the maximum attainable under ideal circumstances. Now for the Zeiss approach. They state that these resolution figures cannot be used to define the image quality of a lens or a photographic system. The human eye can resolve black white line pair when the difference in contrast is about 2%. Such a small difference will be covered by the large image noise in a system (lens/film combination). Zeiss uses as criterion for useful resolution in high resolution photographic emulsions, when the contrast value is above 30%.
|Film||ISO||RMS granularity||Maximum Lp/mm for halftone photography||Lp/mm at 50% contrast|
If we check the manufacturers data for this criterion we get the following results. Contrast value has been set somewaht higher at 50%. To be explained later.
We now see that the figures for maximum resolution are not so relevant. At least they can not be employed for photographic purposes, as for photography we need contrast.
The maximum figures demand diffraction limited lenses for a small spectrum of wavelengths. The table for diffraction limited resolution shows that even if the Agfa Copex could deliver the 600 lp/mm we need a lens with an aperture larger than 1.4 that is diffraction limited. Such a lens does not exist for photographic purposes.
When some reports state that they indeed have succeeded in getting 600+ lp/mm on film with consumer-type photographic optics, then they seem to operate beyond all known limitations of todays optics and visual qualities of the human eye.
The impact of the lens.
The best photographic lenses from Leica can deliver a theoretical resolution of about 450 lp/mm at a 5% contrast. When the limit of 30% is applied, about 250 to 300 lp/mm are possible and indeed in areal projection of test patterns, we can detect this value on the screen. But keep in mind the comment form Zeiss, that this detection is not useful for the definition of image quality.
When we take pictures in the usual situation (with films, enlargers and printing paper, this figure drops another stage. The best I got in a controlled situation, (see below) is 120 lp/mm ON THE NEGATIVE, and enlarged 15 times on a print gives about 8 lp/mm, which in practice is still reduced to about 6 lp/mm as the enlarger lens also reduces the quality a bit.
Kodak had suggested that a photographic lens needs to have a resolution 3 times as high as that of the film to fully exploit the capabilities of the emulsion. There is also a nice equation to relate film and lens resolution to get final or system resolution:
(1/R)2 = (1/R)2 + (1/R)2.
When we apply this equation and Kodak's rule to some examples we get this table.
|Lens resolution||System resolution|
This table does indicate that Kodak's rule holds. A film resolution of 100 lp/mm can become the system resolution when the lens resolution is at least 3 times higher. Most lenses on the photographic market do have realistic resolution values much lower than 100 lp/mm (when the threshold of 30% contrast is applied) and around 100lp/mm when we relax this criterion quite a bit.
So we see that the film resolution of 100 lp/mm, that we established above as practical, given the manufacturer's data and the Zeiss rule of 30% needs a lens of at least 400 lp/mm to be recorded on the negative. And the reverse holds too: a film with a theoretical resolution of 600 lp/mm is hardly better in practice than a film with 100 lp/mm, when a much better lens is used. Given the correction state of most current lenses, it is more true to life to use a lens resolution of 100 as the standard. These figures are in close accordance with my practical findings as my tests with very good Leica lenses gave system resolution with a wide range of films in the 12 to 100 ISO class of 80 to 120 lp/mm.
I started using a high quality microscope with a magnification of 40, 100 and 400 times. Then I used a testchart with lines and circles (as the pattern itself will influence the resolution limit). I set up the Leica in front of the testchart at a distance that gives a negative magnification of 100 times. The idea was that when using the microscope at M=+100, I would be able to see the resolution pattern that is closest to the 600 lp/mm. I used the Apo 135 as this one is capable of resolving at least 300 lp/mm at a acceptable contrast and even 450 lp/mm at a very low contrast (less than 10%).
To ensure optimum results I used of course the center portion of the negative and to make sure film flatness and focussing errors are not a problem Iused the following setup. A Siemens Star was used to check accurate focusing (a phase shift in the pattern will indicate a focus error) and I did extensive focusing bracketing by marking the distance on the lens and taping scaled paper on the mount to accurately make my bracketing. Result one is that the lens and the camera focussed extremely accurately even at 13.5 meters with a 135mm lens, which is reassuring in itself. I shot three films and checked every negative under the microscope to find the best results. As resolution tests always involve errors in viewing, I used the best 10 results and averaged the numbers to get a result that is at least in principle reproduceable by anyone.
APX 25, APX 100, Delta 100, Pan F, TMax100 and Technical Pan were selected. Gigabit film is standard Agfa Copex Microfiche film (no new emulsion this Gigabit) and TP is also a micro-film, which are "forced" to go for continuous tone negatives, I wanted to use films designed from the start as continuous tone film as a comparison.
This study is not designed to be a lens test. The use of several lenses was necessary to ensure that the employment of a specific lens would not unduly impair the validity of the results.
What I tried to establish is system performance and not the absolute performance of a lens or a film. Generally any photographer uses a film and a lens for pictorial or continuous tone pictures and so the absolute?resolution figure of a microfilm (like Gigabit=Agfa Copex and TechPan) of 600 or 400 linepairs/mm is a figure to be interpreted. This figure is based on two assumptions: a film gradient of 3 and the possible resolution of black-white patterns, like letters.
To make these films useful for pictorial photography, the gradient (characteristic curve or CI-value) has to drop to a more normal value of CI=0.6. The Gigabit film just does this and uses a specific developer that gets a CI of 0.5 or even lower. The true speed point of this film is ISO20 and when using it as EI=40 (as recommended) and developing it to a CI of 0.5, we get the classical pattern of underdevelopment and underexposure, good for the dynamic range (overexposure latitude) and resolution, as overexposure will kill any attempt to record very fine details. I used TechPan as a companion film and when developing both films (Gig and TP) in Technidol for the same development time, I got identical results. The much discussed dynamic range of the Gigabit film can be had with the TP too if developed to this effect.
Studying resolution patterns at several densities, I noticed that all these hi-res films are very sensitive to small variations in exposure (plus and minus). Best resolution you get at densities around D=0.4 to D=0.6, which generally is one stop below the grey card value. That may be the reason that hi-res films often are underexposed. You can only compare film capabilities when films are developed to the same CI-value, so my first attempt was to find the correct exposure and development time for all films to be comparable. Having established that I selected the lens for a new series of tests. (My third attempt and many rolls of film used).
Officially best results with a lens are at the infinity position (which is defined optically as the location of the object where the rays are perpendicular to each other and reach the lens at right angles). Normally this at 50 to 100 times the focal length. For all lenses I made tests at a distance of between 5 meter (50mm) and 13.5 meter (135m). Now is 13.5 meter quite inconvenient, so I checked with Leica what would be the closest range without dropping significantly in optical quality. They proposed that 5 meter would be appropriate for the 135mm. The apo-90mm was also used, at 5 meter and 4 meter.
You should realise that films in the speed range of ISO 16, 25 and 40 are difficult to test. At longer distances the flash or the tungsten lamps will be at low power and using shutter speeds of 1/4 sec and an apaerture of 3.4 or 4 is not quite convenient. So I did another thing and ordered new resolution charts (the same Leica are using btw), which make for easier and more accurate testing at the distances to be used. With these new goodies (test patterns and film/exposure/development results for optimal result and comparison), I made extensive series of tests with both lenses at several distances and the following combinations of film-developer. Gigabit in Gigabitdeveloper, in Rodinal, in Technidol and in FX39, TechPan in Rodinal, in Technidol and in FX39. Seven combinations in all, with two lenses at three distances and optimum apertures, including exposure and focus bracketing. That kept me "happy" for a while.? Here I have to include some observations about resolution figures in general.
The difference between 80 lp/mm and 100lp/mm looks big when you interpret in numerically. If you look at the resolution patterns themselves on which the calculation is based, you see a much smaller difference. In other words, there is a law of diminishing returns when you progress in the higher resolution numbers.
My test was setup as follows: the original test patterns have real patterns of 1 and 2 linepairs /mm on the print. They go in steps of 1.1, 1.25, 1.4, 1.6 and 1.8 lp/mm. So if I use a reduction of 40 times and can clearly detect in the subsequent enlargement a pattern of 1 lp/mm and the next pattern (1.1) is blurred, then my film/lens-resolution is 40 lp/mm.
I used reductions of 30 and 40 and 100 times, with the main attention to the 30 and 40 times enlargements: that would still imply a print of 1meter wide from a 35mm negative and who does this on a daily basis.The results?The best resolution I could get (averaged over some observations to account for eye fatigue and some unavoidable subjectivity when deciding if a certain pattern is just resolvable or not): 90 lp/mm in Gigabit with Rodinal (both Agfa products). 90 lp/mm with TechPan in FX39, 80 lp/mm with TechPan in Technidol LC (both Kodak products) 80 lp/mm in APX25 and TX39 80 lp/mm in Tmax100 in FX39.
Every film/developer combination could easily produce 60 lp/mm IF and WHEN developed and exposed and focused accurately. If any of the three variables is off the optimum value, results drop to 40 lp/mm or even worse.Of cOf course you need to address the issue of grain and tonal range and dynamic range etc and here we see that best are TechPan and Gigabit (but only when exposed/developed correctly!), with APX and TMax100 (ex equo) and just a small step behind the hi-res films.
If you do realise that Tmax100 has a speed of ISO100 and the others barely get ISO25, that the tonal rendition of Tmax is excellent and that it has a dynamic range of 7 to 8 stops, when developed appropriately, the overall winner should be Tmax100!
The startling conclusion for me is that there seems to be a threshold of useable resolution of about 60 to 80 lp/mm on the negative. This value is attainable when using outstanding equipment in the right combination of all elements of the imaging chain.
The use of hi-res films will give you an advantage in grain when enlarging beyond 15 times, but you will hardly get better useable resolution. To move beyond this threshold to values of 80 or 90 lp/mm, can be done, but the additional care, accuracy and control over all parameters is extremely demanding and you may question if this additional amount of control and energy is worth the effort.
To be realistic: a true 60 lp/mm with high contrast (= high MTF values) on the 35mm negative will deliver a 15 times print enlargement with a print resolution of 4lp/mm, which is extremely high and needs a very attentive eye to detect and a very high level of expertise to get on the print.
The maximum figures quoted out of context for diffraction limited lenses and high resolving power for micro-films are theoretically true no doubt about that. When translated into general photographic practice, these figures need to be downgraded very much to allow for optical aberrations, system resolution, contrast drop, picture technique, camera vibration etc.
Even allowing for all these effects the goal of 100 lp/mm on the negative is hard to reach. On the other hand, this study has shown that the 100 lp/mm are the limit of what we need (or can use) for high resolution imagery with 35mm equipment.
Most photographers do not care whether they are photographing with 20 or with 80 lp/mm, as they are interested in good content and set high values on art and emotion as parameters for good a good picture.
No trouble at all. But for those who want to know where the limits are, it may be a sobering thought that the jump from 20 to 80 lp/mm, while seemingly a small step, is very difficult to attain. And the proposal that 600 lp/mm are viable on the negative is theoretically and practically illusory.