Leica SL, part 2
ISO speed and dynamic range
Leica does not specify the nominal speed of the sensor and restricts itself to the listing of the range of speeds. This range runs from ISO 50 to ISO 50000, a speed range from 1:1000. It should be well-known that all speeds higher than the nominal one are the result of the combined effect of underexposure and increased gain at the output channel of the analog-digital converter. The gain is simply the strengthening of the signal which also means an increase of the noise related to this signal. The algorithms employed to balance the signal-noise ratio are effective. From ISO 50 to ISO 400 the images have the same detail definition and almost the same noise pattern. At ISO 800 noise increases and detail definition drops. From ISO 1600 to ISO 3200 the images are still useable, but from ISO 6400 till ISO 12600 noise increases sufficiently to wash out details and at the two highest ISO values ISO25000 and ISO50000 only the major outlines of larger subjects are detectible. The illustrations below are sectional enlargements.
Below: ISO 200 no noise reduction
Below: ISO 1600 no noise reduction
Below: ISO 25600 (no noise reduction!)
In practice the ISO setting of 6400 is still useable.
Resolution figures were the main criterion for lens performance in the 1960s. Dynamic range figures have the same status for the performance of the sensors and are equally misguided. The maximum numbers are not interesting. Remember that film emulsions had a density range from 0.0 to 3.6 in steps of 0.3 per full aperture. The useful range however was reduced to 0.15 to 1.35 because in the deep shadows and the highlights no structure could be detected in the negative and after printing. For the check on the dynamic range I relied on the classical test. Take a picture of a grey chart in ambient light and measure the light level with an exposure meter. Then under- and overexpose by 6 stops to see at what step the readings are below 10 and above 250 in digital terms. The accompanying graph indicates that the SL at ISO 200 has a useful range of -4 to +2 stops. This range is also available at ISO 25000 with a slightly increased shadow recording. In the days of slide film the photographer had to be very careful not to over-expose because washed out highlights could spoil the picture and in over-exposed areas no detail or gradation could be detected. This phenomenon is still a problem with digital capture. It is perhaps nice to have a high ISO speed of 25600, but Leica would be well advised to focus attention to the over-exposure problem. The six-stop range of useful dynamic rage may be lower than the theoretical maximum of more than ten stops. The useful range of six stops however may be extended a bit and can match the benchmark of the Zone System of seven stops from bright to deep shadows both with visible textural detail.
Below: ISO 200
Below: ISO 25000
The electronic viewfinder is very bright and is in no ways inferior to the classical glass screen of the SLR with a pentaprism. The brightness however is related to the exposure reading of the scene and represents in effect the brightness of the recorded scene on the sensor, given the selected exposure level. The upshot is that in low ambient light levels the viewfinder shows a much brighter image than what is seen with the naked eye. The viewfinder does not let you see the atmosphere of the scene as a SLR screen or a rangefinder would. It is of course possible to move (in the spotmeter option) the spot to a part of the scene such that the brightness of the EVF represents the mood of the scene. An override would be most useful.
The stabilization function is an effective tool for slow speeds. I lack the necessary tools for an exact analysis, but handheld shooting at speeds from 1/4 to 1/15 second show a level of detail and lack of vibration that would be impossible to achieve without OIS. I have to admit that this function depends on the ability of the user for stable shooting and there is a large range of difference in this ability. I once did a test (log ago) that proved the usefulness of image stabilization. The impact of the stabilization mechanism on image quality is carefully controlled. This control is also (partly) responsible for the number of lens elements.
Below: handheld 1/13 sec
The AF works well for stationary objects and for slowly moving objects in a straight line. When the object is unpredictably moving in 3D space the AF function does not cope with the situation. The phase-detection systems have an advantage here. Focus tracking is one of several options in the Leica SL menu catalogue and on paper has a promising performance Combined with the motor-drive option of continuous shooting the AF function cannot cope and the accuracy drops. The more so when the objects are moving fast and unpredictable. There is a balance between speed of focusing and speed of recording. The two options are precision preference with a lack of speed or speed preference with a lack of precision. A new third option might be the best: let the AF precision govern the recording speed, that is the option between AFc and AFs. When the single shot option is selected the AF is accurate but the tracking option stops after pressing the shutter. One would expect that tracking would continue while taking more shots one at a time. The tracking option would be fine for following a moving object and be able to take more pictures during the following act, for instance to improve composition or pose. Here one would have wished that the Leica engineers and/or product managers had invested more time in improving some useful options to perfection. At least it would be commendable had one strived to offer more functionality beyond what the competition has and what would really be needed by the practicing photographer, who wants to use a Leica camera with features that stand above the crowd.
The performance of the lens is beyond comment. This is very frustrating for any reviewer who wants to have something to criticize. It is of course possible to force the lens to exhibit some secondary reflections. In most situations however the lens is commendably free of flare and secondary reflections. The definition from center to edge at all focal lengths and apertures guarantees crisply rendered fine detail. Contre-jour pictures, always a good test for any lens, produce deep shadows and halo-free highlights. The size of the lens is the consequence of the optimized performance, especially at the closer distances below 0.5 meter. The OIS unit asks for space and so does the AF unit. The AF unit is a ring-shaped device with a stepping motor that controls 760 steps in a 360 degree circle. Every single step moves over a distance of 0.8 micron (less than one thousandth of a millimeter!), an accuracy that not even the mechanically focusing M lenses can beat. The lens is constructed with 18 lens elements and has five aspherical surfaces. This design points to two aspects: (1) the new rule is that more elements give better performance which (2) is a departure from the classical Leica design statement that the minimum number of elements gives the design an edge in performance. Now Leica has joined the ranks of Panasonic and Zeiss (among many others) that an improvement in performance implies the deployment of more and more lens elements.
The close-focus capabilities are one of the hallmarks of this lens as is the overall performance. The new design style is partly responsible for this quality. There is indeed a subtle evolution noticeable from M over S to SL and Q. (More about this change of design can be found in my forthcoming book about Leica optics).
Below: ISO 3200 handheld 75mm f/4
below: contre-jour ISO 200, 90 mm f/4
Below: 50mm at ISO200 f/5.6
SL versus DMR
It is en vogue to state that every new product is better than anything that went before and sets a new standard. It is interesting to compare the SL with the original DMR Module for the R8/9. The sensor size of the DMR is 26.4 x 17.6 mm or 464.64 square mm. The sensor size of the SL is 864 square mm. If we would only use the area of the DMR the SL would have 12.91 Mp. The DMR has 10 Mp and a pixel size of 6.8 micron. The SL has a pixel size of 6.0 micron. I happen to have a few pictures shot with the R9/DMR in 2005. Within the timeframe that the SL was available it was not possible to create pictures that could exactly replicate the original DMR pictures. The pictures below give a fair impression of the differences. The images are converted to BW to eliminate the distraction that color may introduce. The differences are mainly detectable in the very fine structures at greater magnifications. When the prints on A4 scale are studied the finer structures in the skin of the model (the SL picture) may be visible. The main difference is the possibility for enlargement: the DMR image is at its limit when the screen magnification is 130% and the SL image can easily be enlarged to 200% without loss of detail. I have often noted that the number of pixels is more related to the size of the final print and not so much to the quality of the definition. The classical relation between the 135 sized negative and the 120 sized negative still holds: the size of the magnification without loss of definition is higher when the image is shot on 120 film, all other things being equal (ceteris paribus condition!)
Below: DMR, 50 mm f/8
Below: SL 50 mm f/4