Hello,
Concerning the spreadsheet I think it's a quite valuable tool for somebody who is intending to buy a new camera or just plays with the options to determine whether he should buy a new camera. I don't want to repeat the article or previous discussion, however I quite agree with the point that the DxO Mark is a disturbingly partial quality measurement considering that it does not take into account resolution.
Actually in the resolution area I would like to propose an amendment, which helps to assess the actual benefit of a higher resolution sensor given the limitations of required lens quality and diffraction (e.g. see the article 'Understanding Lens Diffraction' on this site). For crop format sensors at and above a roughly estimated 12MPixel only a very good, normally pro quality lens can deliver a matching resolution - and only if not stopped down beyond roughly F8 (due to diffraction) - or for not as good quality lenses additionally only in the central image area. For full format sensors at least the diffraction limit, which is always present regardless of lens quality, is substantially increased in proportion to the increase in physical size of the image as given by the size of the sensor. Consequently a full format sensor is able to deliver substantially higher resolution than a crop sensor using the same lens, provided the resolution of the crop format sensor (or of both sensors) exceeds the diffraction limit and provided a lens quality which avoids excessive resolution loss at the image periphery captured by the full format sensor, but not by the crop sensor.
Additionally, a lens' resolution limit is physically given in line pairs per mm (and similarly most lens imperfections), therefore the no. of resolved line pairs is increased by a full format sensor compared to a crop sensor with the same amount of MPixels, simply because the physical image size (= sensor size) is larger.
Therefore I propose to include also the sensor size in the spreadsheet, in the most simple way to indicate the sensor size. A quantifiable and in my view better indication is to standardise the sensor resolution measured in no. of pixels in sensor height direction by the physical image height, i.e. to multiply the sensor's no. of pixels in sensor height direction with the sensor height (in mm). Effectively, this corresponds to a measure of line pairs per image height.
What do you think?
Hello Adi,
I see you have just joined - welcome - and thanks for sharing your suggestions on this exercise.
You have a number of clearly relevant factors here, so let me try to work through them with you. Firstly, I think you can appreciate that many of these camera bodies can accommodate a huge variety of lenses, each one of which has unique performance characteristics - which change with each f/stop! It would really be a massive exercise to construct a spreadsheet which deals adequately with the resolution of lenses, and as I don't have the time or materials to do that, much as it would be nice, I wouldn't be the one to attempt it.
Secondly, you have correctly identified a key constraint influencing resolution - that being the performance of the lens. It is true that our highest resolution sensors can out-resolve many of the lenses we put over them, so the binding constraint on resolution becomes the lens rather than the sensor. In fact, this was true even six years ago when Canon produced the first 11 MP professional DSLR. Their cheaper lenses showed their limitations quite clearly on this camera body. So it is even more pronounced nowadays.
Thirdly, it is in general terms correct that lenses perform better in the center than at the edges, so if you are using a camera with a full-frame sensor you would get more of the lens' less desirable edge effects than if you were using a "crop-sensor" where a larger proportion of coverage is from the higher quality portion of the projected image. But that doesn't make or break the desirability of using a full-frame sensor - it speaks more to the desirability of having and using better lenses. The resolution of the sensor and the resolution of the lens are independent variables governed by different factors. In the case of the sensor, one of the main issues affecting resolution and image quality in general is the pixel density - the higher it is (i.e. the more photosites that are crammed on to a given space) the noisier the sensor is likely to be regardless of what lens you put over it. This characteristic of sensor design - the relationship of pixel count to sensor size and its effects - seems to be well-captured by DxOMark, so I think we shouldn't add variables for this - it would appear to be redundant. In fact, when I started this exercise, the first draft did include quite closely what you are suggesting - a metric of pixel density as a quality indicator, but then part way through the preparation, DxOMark came along with a much more far-reaching approach which includes it, so I scrapped that and went with DxOMark.
In the case of lenses, as you have noted, and by the way also explained quite thoroughly at
Cambridgeincolour the aperture has a major impact on diffraction which in turn impacts resolution. So indeed, you can spend huge amounts on the highest resolution sensor available, and undo some of its potential by using poor quality lenses or using good lenses sub-optimally. The best one can do about this is to try using the lens at its optimal aperture as much as circumstances allow, and this situation/advice applies regardless of the sensor.
So conveying this into an approach for deciding on a camera-lens combination, rather than trying to construct and interpret large complex databases for the multitude of sensor/lens combinations and lens apertures one can imagine, the more tractable approach would be to focus first on the camera body and optimize that, resulting say in a short-list of best candidates for your needs - then look at the MTF charts available for the various lenses of interest to you and useable with the bodies on your shortlist. If your objective is to maximize resolution and image quality, work back and forth between the camera bodies and the lens data to see which combinations are each best in their own right relative to your needs, and compatible with eachother.
Hope this helps.
Mark