Leica M9, part 5, noise and dynamic range.
Introduction
For this test I used the OECF noise chart with 20 patches covering a 10.000:1 luminance range. This is the industry standard. The luminance range corresponds to a range of more than 13 stops. The graphs I will present are mathematical analyses of the image files, made at different ISO values. They may look a bit intimidating at first glance, so here is a short explanation. OECF stands for Opto-Electronic Conversion Function and is widely regarded as the best determinant for dynamic range analysis.The top left graph (the string of grey squares) on the image is this OECF and it may be interpreted as the classical density (characteristic) curve of a film emulsion, starting with a low density (shadow) at the bottom left corner and ending with a high density (highlight) at the top right corner. Films follow the well-known S-shaped curve and you see this shape again in the graph. The bottom graph represents the noise level, in the same way as grain size was measured in the AgX world. In the digital world the Signal to Noise Ratio is the best known parameter, denoted as SNR. This parameter is measured in dB and the formula is SNR = 20 x Log (10) (Vs/Vn). The practical range goes from zero to 60 dB, with the value of 20 dB as the watershed value; higher values go from acceptable to excellent and lower values go from not acceptable to awful.
This scale is not the same in all graphs but is dynamically and logarithmically adjusted. The yellow horizontal bar in the middle graph represents the noise level where good quality images start.
The figures on the top right side of the set of graphs shoe the dynamic range in maximum values and for different quality levels. As example: the D3X at ISO2500 has a maximum dynamic range of 13 stops, but this is theory. Several patches have the same density and are indistinguishable. This is the same situation as with film where we can have a very high density that cannot be brought to print with just noticeable luminance differences. This state of affairs is the reason why Ansel Adams, designing the Zone System, restricted the range to ten zones, each one just measurable and visually different from the next one.
The D3X at ISO2500 can deliver high quality imagery when the number of zones is limited to 5.6 stops, close to what a good slide film can handle. When a lower quality can be accepted the D3X can cope with ten stops or the full range of the Zone System. In itself not a bad result, but one to accept with restrictions.
The cameras were all set for manual white balance to correspond with the color temperature of the test equipment. Note that the Nikon and M8 have excellent grey values, where the M9 is blue biased. In this area Leica needs to do some additional home work. I did not adjust the color balance as this is not the goal of this part of the range of M9 articles. For the record: the source light has a Kelvin value of K5000 and all three cameras were manually set to this value. A check with Capture One to see if the cameras recorded this source to correct values gave this result
D3X set to K5000, recorded as K5000
M8 set to K5000, recorded as K5000
M9, set to K5000, recorded as K5450
ISO 160 on D3X, M9 and M8
The D3X has a useful dynamic range of ten stops and an SNR of on average dB 47. The M8 has a useful dynamic range of eight stops and an SNR of on average dB 47. The M9 has a useful dynamic range of seven stops and an SNR of on average dB 43.5. The M9 is remarkably somewhat less good than the M8.2 I used. The differences are small however, but visually present. The Nikon is not better and this is surprising as the Nikon D3x has been championed as a speed king. But at lower ISO values this advantage is not clearly visible.
The D3X has a useful dynamic range of ten stops (but for maximum quality the range is only 6 stops) and an SNR value of on average dB 47. The M8 has a useful dynamic range of eight stops (for maximum quality it is six stops) and an SNR of on average dB 46, almost equal to that of the Nikon camera. The M9 has a useful dynamic range of seven stops (for maximum quality it is six stops) and an SNR of on average dB 43, which is lower than that of the other two cameras, but not by much. I cannot help but warning that to attach too much value to small numerical differences is as dangerous as relying too much on subjective visual interpretation. I have always been skeptical of the many magazines that rank cameras and lenses in percentage points: a camera/lens combo with 73% is always better than a camera/lens combo that scores 72%. This is nonsense! The D3X has a useful dynamic range of ten stops (but for maximum quality the range is just below 6 stops) and an SNR value of dB 42 on average. The M8 has a useful dynamic range of eight stops (for maximum quality it is just below six stops) and an SNR of on average dB 43 a slightly better result than that of the Nikon camera. The M9 has a useful dynamic range of seven stops (for maximum quality it is just below 6 stops) and an SNR of on average dB 41, just below the level of the other two cameras. The consistent result by now is that the D3X is best, closely followed by the M8 and the M9 is always a bit behind the M8.
ISO 1250 on D3X, M9 and M8.
The D3X has a useful dynamic range of ten stops (but for maximum quality the range is just below 6 stops) and an SNR of on average dB 45. The M8 has a useful dynamic range of eight stops (for maximum quality it is just above 4 stops!) and an SNR of on average dB 40, below that of the Nikon camera. The M9 has a useful dynamic range of seven stops (for maximum quality it is just below 4 stops) and an SNR of on average dB 39 which is below the level of the other two cameras. From now on the Nikon camera forges ahead with the ISO speeds and noise reduction. The D3X has a useful dynamic range of ten stops (but for maximum quality the range is just below 6 stops) and an SNR of on average dB 40. The M8 has a useful dynamic range of six and a half stops (for maximum quality it is just above 3 stops!) and an SNR of on average dB 36, below that of the Nikon camera. The M9 has equal values.
Conclusion.
It is clear that the Nikon has the edge, a small one up to ISO 320, a bigger one up to ISO 640 and a major one from ISO 1250. The ISO 2500 is not really useable on the Leica M8/9 camera, but as a last desperate resort is is an option. There is a persistent tendency to hype up the performance of the Nikon high end digital SLR cameras as far as noise and dynamic range at higher ISO values is concerned. This is true above ISO 1250, but at lower values the competitive advantage compared to the Leica M8/9 is not very pronounced, al be it visible. For me an excellent ISO 400 setting is a great improvement. just now that Kodak has released the very good T-Max 400 film. Generally one may note that for best quality at higher ISO values a dynamic range of four to five stops is not a formidable achievement. It is interesting to reflect on the fact that the M8 performance is just ahead of the M9 performance. Differences are not impressive, but just there. The final result that the M8/9 can keep up with the formidable Nikon up to ISO400 (to be really critical) is a pleasant surprise. Of course one can comment that the Japanese high end dSLR cameras are excellent up to ISO3200 (with the appropriate post processing), but I am happy with a very good ISO 400 and excellent wide aperture lenses. This combination is well suited for the goal of emulating high quality reportage and documentary photography in film-look-alike. And that is what Leica is famous for.