What do Thalmann’s lens tests tell us about sharpness, and how do they connect to MTF theory and the quiet power of large format film?
Factors behind Thalmann’s Future Classic lens selection
In my previous post, I examined the factors that influence Thalmann’s Future Classic lens selections. One major factor I identified was weight—that is, portability played a significant role.
An open question was whether MTF specifications or resolution performance influenced Thalmann’s selections. To clarify this, I reached out directly to Tim Parkin, who wrote an insightful article in On Landscape discussing Thalmann’s Future Classics list.
He confirmed that Thalmann’s choices were primarily driven by a balance between weight, image circle, and resolution, with weight being particularly important for landscape photographers, as Thalmann himself was. Tim also pointed me to the well-known Hevanet test, a collaborative effort between Kerry Thalmann, Christopher Perez, and Mike McDonald (circa 1999). Hereafter, this test will be referred to as the Thalmann Test for simplicity. This test provides additional insight into the performance characteristics that likely informed their evaluations.
Thalmann’s Lens Test
Thalmann and collaborators measured the resolution of several lenses, evaluated together with film at specific tested apertures. Each entry in the corresponding column lists the aperture (f-stop) at which the test was conducted, followed by three resolution values. These values represent the measured resolving power at the center, midpoint, and corner of the image field, respectively. All results are expressed in line pairs per millimeter (lp/mm), a standard unit of spatial resolution, where higher values indicate better performance. For example, an entry of f/16: 60 54 32 means that at f/16, the lens achieved 60 lp/mm at the center, 54 lp/mm at the midpoint, and 32 lp/mm at the corner.
According to additional details shared on Christopher Perez’s blog, all lenses in the test were evaluated at a 1:20 magnification ratio, representing a typical working distance for large format photography. The test setup used TMax 100 film, developed in D-76 according to a calibrated system of exposure time and temperature. A variety of 4×5 cameras were employed, including the Canham DCLH, Linhof Technika III, Linhof Master Technika, Linhof Master Kardan, and the Tachihara wood field camera. The Edmund Scientific Lens Resolution Chart was used for measurement. Exposures were carefully controlled, targeting Zone VII–VIII for detailed highlights and Zone III for detailed shadows, ensuring consistent tonal evaluation across all negatives.
Based on this description, it becomes clear that Perez and Thalmann did not measure full Modulation Transfer Functions (MTFs), but rather reported the combined resolution of the lens-film system. The distribution (histogram) of these measured resolutions, taken at the center of the lens at f/22 (as an example), is shown in Figure 1. The measured resolutions range from 35 to 75 line pairs per millimeter (lp/mm), with the most frequent value (mode) occurring at 60 lp/mm—a range where Schneider and Rodenstock lenses are particularly well represented. Fuji measurements tend to cluster in the mid-resolution range, while Nikon entries are spread more evenly across the spectrum. Very high resolution values above 70 lp/mm are rare, with only a single Rodenstock entry reaching that level. The chart highlights Schneider’s strong and consistent performance across the higher-resolution bins, suggesting robust resolving capabilities under the testing conditions applied by Perez and Thalmann.
To assess whether Thalmann’s selection of “Future Classics” was based primarily on lens resolution, the measured data for lens-film systems were grouped into two categories: Future Classics (in red) and the original Perez-Thalmann-Test dataset (in blue). If resolution alone had strongly influenced the selection, we would expect the red group to be noticeably shifted toward higher resolution values. However, as shown in Figure 2, many of the Future Classics fall within the same resolution bins as the more frequently occurring entries from the Perez-Thalmann-Test, particularly at 50 and 60 lp/mm. This indicates that the Future Classics exhibit resolution performance comparable to historically well-regarded lenses. Notably, very few Future Classics exceed 65 lp/mm, suggesting that ultra-high resolution was not a dominant selection criterion. Instead, the data imply that Thalmann’s choices were likely informed by broader criteria—possibly including build quality, rendering characteristics, or reputation—rather than resolution alone. Thus, it can be concluded that measured resolution was not the primary factor in designating a lens as a Future Classic.
Finally, the relation between listed price and measured resolution is depicted in Figure 3. The graph demonstrates that there is no strong correlation between lens price and measured lens-film resolution among large-format lenses. While some high-priced lenses achieve excellent resolution, others do not, and several mid-range lenses perform remarkably well. The downward trendline indicates that a higher price does not necessarily correspond to better optical performance. Notable cases include the Xenar and G-Claron, which deliver high resolution at relatively low cost, and the Super-Angulon, which commands a high price despite modest sharpness. Overall, this suggests that lens sharpness is not a reliable predictor of market price in this category. The price data used were sourced from the MKD45 database.
The Unresolved Puzzle
Once this conundrum on the main selection criteria was resolved, two key questions remain:
- Can the compound resolution measurements from the Perez–Thalmann tests be estimated from the MTF curves published by the manufacturers?
- If yes, which MTF(%) represents the lens resolution reported in this test?
It’s essential to note that only Rodenstock and Schneider-Kreuznach provide publicly available MTF data; Nikon and Fujinon do not publish comparable curves, which limits broader comparisons.
Before attempting to answer these questions, I’ll briefly explain what MTF (Modulation Transfer Function) is, how to combine the MTFs of a lens and film, and how to estimate average lens sharpness to allow for meaningful comparisons across different models.
MTF in a nutshell
Modulation Transfer Function (MTF) describes how well an optical system, such as a lens, film, or scanner, preserves contrast at different spatial frequencies, typically measured in line pairs per millimeter (lp/mm). It quantifies the system’s ability to reproduce fine details, with MTF values ranging from 0 (no contrast) to 1 (perfect contrast). In essence, the MTF curve reflects how sharply and accurately various levels of detail are rendered in the final image. A comprehensive explanation of the theory behind MTF is available in several books and links, and will not be repeated in detail.
In essence, every optical system—including the human eye—has its own Modulation Transfer Function (MTF). When multiple components are part of the imaging chain, their individual MTFs combine multiplicatively. For example, in a system consisting of a lens, film, and scanner, the overall MTF is the product of the MTFs of each component:
\[ \text{MTF}_{Total}= \text{MTF}_{Film} \times \text{MTF}_{Lens} \times \text{MTF}_{Scanner} \]
Estimating MTF values of a Film + Lens System
Norman Koren developed a MATLAB script for this purpose. Since I don’t use MATLAB and prefer open-source tools, I translated the code into Python for my use. The Python version (mtf.py) is available on GitHub for anyone interested in exploring or working with MTF data.
For example, if we want to estimate the MTF for TMax 100 film used with a 150mm f/5.6 Schneider Apo-Symmar lens focused at infinite and f/22, the script generates the MTF curves shown in Figure 4. I am using the same example and notation as Koren to make the comparison easier to follow. At this point, a brief explanation of the underlying theory is necessary to follow the discussion.
- Distance from the image center (field position)
- Aperture (f-stop)
- Wavelength or light spectrum
- Focal length
- Focusing distance
- Orientation, typically measured in two directions: Sagittal (along the radial direction from the center) , Tangential (perpendicular to sagittal, following a circular path around the image center).
For these reasons, lens manufacturers seldom publish MTF graphs showing contrast (MTF) versus spatial frequency, as seen in Figure 4, for each lens. Instead, they typically provide a series of plots for each lens model that represent MTF performance at several magnification ratios and aperture settings. Each plot is a 2D graph where the x-axis represents the image height (i.e., distance from the optical center), and the y-axis represents contrast (MTF). Within each plot, separate curves are drawn for several fixed spatial frequencies — typically 5, 10, and 20 line pairs per millimeter (lp/mm) in the case of large format lenses. These represent coarse contrast, mid-level detail, and fine resolution, respectively. For each frequency, two curves are usually plotted: one for the radial (sagittal) direction and one for the tangential (meridional) direction, allowing users to assess astigmatism and off-axis performance. See, for example, an MTF chart, for given parameters) published by Schneider-Kreuznach for this lens in Figure 5. This graph was retrieved from the Internet Web Archive.
In a perfect lens, the MTF curve would be smooth and ideal: high contrast at low spatial frequencies, gradually decreasing to zero at the diffraction limit. The curve would be the same in both radial (sagittal) and tangential directions, meaning no astigmatism or field-dependent aberrations. It would also be flat across the image field, maintaining identical performance from center to edge. In essence, all MTF curves would overlap cleanly, showing perfect symmetry, no falloff, and a graceful decline governed only by diffraction, not by lens imperfections.
MTF of films
Using the MTF plots provided by manufacturers, we can model the overall system performance by applying some straightforward mathematics to both the film and the lens components.
To introduce the concept, let’s begin with the MTF of the film itself (see Figure 6). The shape of a typical film MTF curve can be approximated using a Lorentzian function, which is defined by two key parameters: the characteristic frequency \( f_{50} \), and the falloff rate \( \gamma \). The equation is as follows:
\[ \text{MTF}_{\text{film}}(f) = \frac{1}{ 1 + \left( \frac{f}{f_{50}} \right)^{\gamma}} \]
The MTF of Kodak Ektachrome E100VS was retrieved from the Web Archive.
Here, \( f_{50} \) represents the spatial frequency at which the MTF drops to 0.5 (50%), while \( \gamma \) is an empirical exponent that controls how sharply contrast falls with increasing detail — typically ranging between 1.5 and 2.5. For T-Max 100 film, \( \gamma = 2.24 \), value chosen to accurately fit measured data and ensure the function behaves well when inverted.
Green is typically the most relevant wavelength in resolution analysis because the human eye is most sensitive to it — far more than to red or blue. In the example shown in Figure 6, the \( f_{50} \) value corresponding to green light is approximately 41 line pairs per millimeter.
The parameter \( f_{50} \) is one of the most meaningful indicators of real-world image sharpness. Unlike theoretical resolution limits defined at 10% or 100% contrast, \( f_{50} \) represents a practical midpoint that closely matches how the human eye perceives detail and contrast. It provides a stable, comparable metric across different lenses and films, and is especially valuable when modeling overall system performance or estimating usable detail in final prints. In essence,it captures the critical balance between resolution and contrast that defines visual sharpness. \( f_{10} \), on the other hand, it’s often used as a practical indicator of a system’s theoretical resolution limit — that is, the finest detail the lens or film can still reproduce with minimal but usable contrast. This metric will be essential in addressing the question raised earlier, as it defines the boundary between visible detail and theoretical resolution.
For photographic films, a MTF is generally isotropic — that is, the same in all directions. However, the development process, particularly for black and white films, can significantly influence the MTF curve. Factors like developer type, dilution, agitation, and temperature all affect contrast and grain structure. Table 1 lists typical \( f_{50} \) values for a range of known films.
As shown in Table 1, T-MAX 100 stands out with a remarkable \( f_{50} \) value of 125 lp/mm, marking it as an exceptionally high-resolution film. This outstanding performance was a key reason it was selected for the lens tests conducted by Perez and Thalmann, mentioned above.
MTF of the lens
The same equation used to model film MTF can be adapted to describe lens performance. To do this, we simply replace \( f_{50} \) with \(f_{\text{lens}} \), following the notation introduced by N. Koren in his blog. Here, \(f_{\text{lens}} \) represents the spatial frequency at which the lens MTF drops to 0.5, or 50% contrast.
The value of \( f_{\text{lens}} \) can be estimated directly from the MTF data provided by the manufacturer. The procedure is straightforward: starting from the Lorentzian equation adapted for lenses, we simply solve for \( f_{\text{lens}} \) as the unknown variable. This yields the following expression:
\[ f_{\text{lens}} = \frac{f}{ \left( \frac{1}{ \text{MTF}_{f}} -1 \right) ^ {\frac{1}{\gamma}} } \]
The values of \( \text{MTF}_{f} \) and \( f \) can be extracted from the manufacturer’s published MTF charts, as shown in Figure 5. As a general rule of thumb, it’s best to use the MTF curve closest to the diffraction limit — typically the one plotted at the smallest aperture. For the Schneider-Kreuznach 150 mm lens discussed earlier, this corresponds to the panel for f/22 at infinity focus. In the case of the Perez–Thalmann tests, the appropriate panel is the one that matches the test’s magnification factor. For simplicity, I will take the MTF values at the center of the image field.
Now, which spatial frequency should we use? For large format lenses, 20 line pairs per millimeter is often considered a representative measure of sharpness. Therefore, we choose \( f = 20 \) lp/mm. In the example shown in Figure 5, this gives \( \text{MTF}_{20} \approx 66\% \). This yields \( f_{\text{lens}} = 28 \) lp/mm for this particular lens.
Once the \( f_{\text{lens}} \) parameter has been estimated for a given lens, the modulation transfer function of the complete system — lens plus film — denoted \( \text{MTF}_{Total} \), can be calculated using the equation provided above.
Summary of MTF Findings
The probability density distribution of the estimated \( \text{MTF}(f_{\text{test}}) \) values — for the 17 lenses with available MTF data — is shown in Figure 7. The mean MTF is approximately 0.13 ± 0.03. Based on this result, and considering the known limitations and statistical uncertainties of the Perez–Thalmann measurements, it can be concluded that the reported resolution values align closely with the \( \text{MTF}_{10} \) level (10% contrast). This level corresponds to the system’s theoretical resolution limit — the finest detail that can still be captured with minimal but perceptible contrast.
This finding is consistent with the stated goals of Thalmann and collaborators. It suggests that lenses performing near the 10% MTF level can be regarded as operating close to their practical resolution limits, offering a useful reference point when comparing MTF data to real-world photographic performance. What these tests, both in general and in the specific example, demonstrate is that film is not the limiting factor for image sharpness. As shown in Figure 4, it has a limited impact on determining overall resolution.
Implications for real-world sharpness
In the case of the Schneider-Kreuznach 150 mm lens used in the previous example, the characteristic frequency can be estimated as \( f_{\text{lens}} = f_{50} = 28 \) lp/mm, focused at infinity. By comparison, a high-quality 35mm lens designed near the diffraction limit typically achieves \( f_{\text{lens}} = f_{50} = 37 \) lp/mm (See Koren). At first glance, this might suggest that the large format lens is less sharp than its 35mm counterpart. But is that truly the case? To answer this, we need to estimate the total detail each system can deliver, and then compare their ratios using the following approach:
\[ \text{Relative Total Detail} = \frac{L_{\text{LF}} \times f_{50,\text{LF}}}{L_{\text{35mm}} \times f_{50,\text{35mm}}} \]
Here, \( L_{\text{format}} \) denotes the linear size (or scaling factor) of the format or sensor, and \(f_{50,\text{format}} = f_{\text{lens},\text{format}} \) represents the spatial frequency at which the lens MTF drops to 50% contrast for that specific format. In the case of 4×5 film compared with 35mm film (cropped to the same aspect ratio), the linear format factors are 4 and 1, respectively, since 4×5 covers approximately four times the image height of 35mm. In this example,
\[ \text{Relative Total Detail} = \frac{4 \times 28}{1 \times 37} \approx 3.03 \]
This implies that a 4×5 image can resolve roughly three times the linear detail of a 35mm system, assuming both are used with good technique: excellent lenses and film, optimal aperture, precise focus, solid camera support, and favorable atmospheric conditions.
This demonstrates that even if the \( f_{\text{lens}} \) of a large format lens is numerically lower than that of high-end 35mm optics, large format lenses still deliver significantly more image information — not due to higher lp/mm performance, but because of the far greater area of the film. This enables greater enlargement with finer detail retention, a fact visible in the comparisons presented by R. Clark. In real-world terms, resolution is not just about sharpness per millimeter, but about the total area over which that sharpness is recorded.
Conclusions
This analysis set out to understand whether the lens resolution values reported in the Perez–Thalmann tests can be meaningfully interpreted through the lens of MTF (Modulation Transfer Function) theory, and what that tells us about sharpness in large format photography.
By estimating the MTF at the spatial frequencies reported in the tests, we found that most lenses cluster around \( \text{MTF} \approx 0.13 \). This suggests that the reported “resolution” corresponds closely to the point where system contrast falls to about 10–15% — a region commonly used in optics as the threshold of practical resolving power. In other words, the lenses in the test were being evaluated near their performance limits, and the results reflect real-world usability rather than idealized lab numbers.
Importantly, we also saw that large-format lenses, despite having lower \( f_{\text{lens}} \) values than some excellent 35mm lenses, can deliver significantly more total detail due to the much larger film area. This becomes clear when we compute the relative total detail — a product of format size and lens MTF performance — showing that a 4×5 image may contain up to three times more linear detail than a comparable 35mm image.
Finally, these results demonstrated that resolution was not a primary factor in Thalmann’s selection of “Future Classic” lenses. As shown in Figure 2, many of the selected lenses fall within the same resolution range as others that were not included. This suggests that Thalmann prioritized a broader set of qualities — including portability, image circle coverage, and rendering characteristics — rather than maximizing sharpness. His choices reflect a thoughtful balance of field usability and optical performance, suited to the practical demands of large-format landscape photography.
