Leica lenses and digital negatives (2007)
When Olympus introduced the 4/3 digital capture system, they made a strong case for the need for new lens designs that could be optimized for the additional requirements for digital capture. Two issues were mentioned in particular: the need for telecentric designs and the need for high resolution/ high contrast lenses. They also claimed that the new 4/3 format would become the standard format for the digital slr domain, because it is smaller in size, enabling higher quality optics in a compact format and because the dimensions 4:3 would fit modern viewers better than the classical 3:2 format of 35mm film. The argument is and was convincing and I even posted the question whether the new 4/3 format could become the (new) digital version of the (classical) Barnack format. This proposition was a bridge too far. Most D-SLR models opted for a APS-C version. The original APS film format was 25.1 x 16.7 mm with an aspect ratio of 3:2 (the classical 35mm format). APS-C sensor sizes vary a bit from 20.7 x 13.8 to 28.7 x 19.1. The bigger sizes are known as APS-H and offer a 1.3 reduction in field of view. There are several reasons for the choice for a smaller format: price of the sensor (a 24x36mm sensor size is extremely expensive), smaller image files can be saved and manipulated faster and stored on memory cards with less capacity (cost effectiveness), and the reduction in vignetting and loss of definition with many lenses when compared to the performance of the same lenses on silver halide media for 35mm photography.
It has been established that lenses designed for 35mm film sizes have problems with vignetting and definition when they are required to record scenes on CCD and CMOS sensors with the same size. This problem is restricted to wide angle and some standard lenses. Nowadays most camera operators use zoomlenses with an extended zoom range starting with wide angle and so many photographers encounter this problem. (Software can lighten the burden and most professionals do not complain). The microlenses on the individual pixels of the sensors were not adjusted to the oblique angles of the rays at the edges of the image circle, specifically for wide angle and standard lenses. Current sensor technology seems to have reduced this problem. At least Canon and Nikon feel confident enough to introduce 35mm sized sensor areas in their new professional cameras. The Canon 1 Ds and the more affordable 5D were the frontrunners of this trend.
There are a number of advantages when using smaller sensor areas, but there is also one very big disadvantage: the definition of the lens needs to be improved. There is a simple and time honoured relationship between resolution and size. The smaller the size of the negative, the better the lens has to be for the same print size! The important point is the print size.
We know from the classical debate about the advantages of medium format and 35mm format that the main issue is the size of the negative. A medium format camera can work with a lens that is optically inferior to a lens designed for 35mm sizes because the enlargement factor of the medium format negative is less when we are considering the same print size. It is a simple case: when we have to capture a certain amount of spatial information on a small area, we need a better lens to differentiate the small details. The classical example is the comparison between 35mm and medium format for a print size of A3: that is 30 by 40cm. A 35mm negative needs an enlargement of 12 times and a medium format negative needs an enlargement of about seven times. The lens for the medium format camera can operate at half the number of frequencies to reproduce the same information content on the print. Where the 35mm lens requires a good MTF value for 40lp/mm, the medium format lens can do with that same value for 20 lp/mm. High resolution in itself is not very informative. A new lens designed for a mobile phone has an MTF of 80% for a spatial resolution of 80 lp/mm. This sounds very impressive and it is! But when we relate these facts to the 35mm size, the factual resolution is 80% for 10 lp/mm and then it is quite normal performance!
Leica M lenses and the crop factor of the M8
Leica lenses have been designed for 35mm negative sizes and silver halide emulsions. The optical designs have been optimized for the frequency range between 5 and 40 lp/mm where a high contrast transfer can be expected. Many modern Leica designs are diffracted limited at medium apertures and can record fine detail of 100 lp/mm and even above this value. You need microfilms, extremely stable tripods, focus bracketing and more to get at this performance. Handheld photography with medium speed films and moving objects (street scenes) is limited to at most 20 lp/mm and often less. The 40 lp/mm limit was established by Zeiss researchers as a practical limit for high definition 35mm photography. My own tests have indicated that is not easy to reach this limit in normal situations. A small focus mismatch or a camera vibration is enough to destroy any ambition for high definition images.
The Leica M8 has a sensor area that is 1.3 smaller than that of filmloading M-models. The lenses on the camera are identical. The rules state that for the same final print size (A4 or A3) the digital negatives of the M8 need to be enlarged by a factor of 1.3 to get the same print size as the silver halide negatives require. And to reproduce the same spatial information content the lens needs to record the image details at a higher frequency: 52 lp/mm at the same level of contrast.
The crop factor does not change the properties of the lens: it functions as a reduction mask on the negative size. If I take a picture with a 50mm lens from a distance of 2 meter, all details of the scene are reduced by a factor of about 40. The M7 negative will capture more of the scene than the M8, but the reduction factor does not change. I can give the M7 an added advantage by closing in to 1.5 meter to capture the same scene content. Then the details are recorded at a lower reduction factor and this will help to reproduce them with good quality.
The imaging chain, based the silver halide process, will progressively reduce the final image quality, where the imaging chain based on manipulation of the image file by post processing software will be able to enhance the image quality. But the software cannot extract details of the scene that are not at all recorded in the first place.
New MTF graphs
The argument as presented above leads to the conclusion that the Leica lenses on the M8 need a good spatial resolution at 52 lp/mm to record and reproduce the same details as the M7 does on film. The Nyquist frequency of the M8 sensor is fixed at about 70 lp/mm. The MTF for the range of frequencies between 50 and 70 lp/mm is of particular interest for this analysis. Several Leica lenses were measured on the bench first for the classical 10, 20 and 40 lp/mm (the M7 format) and then for the new 16, 32 and 64 lp/mm (for the M8 format).
There are some optical constants at work here: the same lens cannot record higher frequencies with a higher contrast than lower frequencies. The Summicron 50mm at f/2 (k=2) records the 40 lp/mm on axis with 50% contrast. The 60 lp/mm must have a lower value. But we are also interested in the performance stopped down. What is the image quality for 60 lp/mm stopped down?
We will analyze the classical Summicron 2/50 and the new Summilux 1.4/50 ASPH and the Apo-Summicron 2/90 ASPH to give the reader an idea of what is possible with current Leica lenses. Here you will find unique graphs of MTF curves. One series is the classical MTF representation for 35mm format with the full image circle of 21.6mm radius. The second series is the MTF representation for the M8 image circle of 16mm radius and MTF values for 16 to 64 lp/mm. The graphs are for full aperture, f/2.8 and f/5.6 for the 50mm lenses and full aperture, f/4 and f/8 for the 90mm lens.
The Summicron at f/2 for the 35mm format shows excellent behaviour for the 10 and 20 lp/mm over the whole image field and even the 40 lp/mm stay above 40% over a large part of the image circle up to 15mm radius. In the corners and edges the definition will drop. The same lens on the M8 delivers excellent performance for the 16 and 32 lp/mm and a drop in contrast for the 64 lp/mm where the edges and corners fall below 20% contrast.
Left is 40 lp/mm and right is 64 lp/mm
The Summicron at f/2.8 adds more punch to the image with an improvement in contrast at all frequencies and a big improvement for the image circle from center to 10 mm radius. The same fingerprint we observe for the M8 sensor: contrast for the 64 lp/mm stays above 40% for a larger part of the image circle.
The Summicron at f/5.6 is at its maximum performance and for the 35mm negative we see outstandingly good image quality over the whole image field.The M8 user may be very happy as the 64 lp/mm is now recorded at a contrast of above 60% with only a small drop in definition at the edges. We mat expect very crisp imagery as the sagittal and tangential lines are very close.
The Summilux at f/1.4 is almost as good as the Summicron at f/2 and the 35mm negative is filled with crisp detail from centre to edge, with the corners a bit lagging. The M8 user can expect the same quality, but now the 64 lp/mm are a bit soft over a larger part of the image area. Here we see the optical limits for this wide aperture for the definition of exceedingly fine detail. We may assume that the Summilux at full aperture will not be used for the recording of such fine detail. Some stopping down will always be considered.
The Summilux at f/2.8 improves upon the Summicron at the same aperture: performance in the outer zonal areas of the negative is significantly better. It is evident that Leica pulled alls tops with the design of this lens. The M8 user will be happy too: only the outer zonal areas of the sensor show a softness that can be improved by post processing.
The Summilux at f/5.6 delivers impressive imagery and only the outer corners and edges stay on the soft side. The M8 user now benefits from the smaller capture area: the softer outer zonal circle is cut off and the image is very crisp over the whole image area with 60% contrast for the 64 lp/mm over the full digital negative area. This performance is extremely good for a lns designed for 35mm photography.
The Apo-Summicron 90mm at f/2 brings the best performance at this aperture for the M8 user: a very crisp recording of the 64 lines frequency over the whole image circle: a straight line from centre to edge.
The Apo-Summicron 90mm at f/4 shows a big improvement in overall contrast and the 64-frequency is recorded with 60% contrast, even better than the Summilux at 5.6. This is obvious the optimum aperture: at f/5.6 contrast drops a little, but the 64-frequency becomes a bit softer over the whole digital negative area: the effects of the diffraction is visible.
The Leica lenses analyzed are computed for 35mm silver halide photography. The MTF results indicate that these lenses perform equally well when put on the M8 with its higher demands for high contrast definition of fine detail.
At full aperture the 50mm lenses are operating at the edge of what is required. The centre performance is beyond reproach, but at the edges the critical limit is reached. Stopped down however, the new Summilux 50 and the Apo-Summicron 90 are very capable performers even at 64 lp/mm which is close to the Nyquist frequency.
The Olympus claim that only lenses designed for the specific requirements of digital recording can deliver superior imagery is not supported by this analysis. The proposal that a future M9 needs a sensor with the same size, but a higher number of pixels would not bring a big advantage. The Nikon approach (large sensor area and large pixel sizes to improve the signal to noise ratio might be a more sensible solution). The quality of the information content (delineation of major subject outlines and reproduction of very fine textural details) is already very good. Assuming that the digital negatives will be printed with high-quality inkjet printers like the Epson 3800, more details are not needed as the printer resolution sets the limits.
With high sensitivities of the sensor and a good signal to noise ratio we can use high shutter speeds with the wide apertures. This will ensure better results that a slow shutter speed and very small pixel sizes where the recording limit is already reached.