Originally posted by johnyates I would think focussing manually with magnified live view would be your best chance to ensure consistent accurate focus.
I have a split screen from focusingscreen, on a KM. The KM is not the best viewfinder, but a split screen is pin point accurate. You have to verify its shimmed correctly by shooting a ruler at a 45 degree angle, running away from you, and confirm that the visual focus corresponds to the "chemical focus"...this is an issue from the darkroom too, where light our eyes are sensitive to is not the same as light the camera is looking at. I dont think the KM does live view. I dont use it for anything other than digitizing negatives and lens testing
We have vision which peaks in the yellow/green area. If you want to do this all really accurately, you would use a beautiful old MX with a great big viewfinder, a yellow green Hoya X0 filter on the lens, and ADOX film with resolution which is higher than the lens. A yellow green filter ensures the film is being exposed to what your eye is sensitive to. You could do the filter on digital too, but it would muck around with the demosaicing, I think.
---------- Post added 27th Feb 2014 at 20:31 ----------
Originally posted by kh1234567890 Is there a link to your test method anywhere ? Sounds interesting.
I can post a link later when I have made a new and improved slant edge...rather than using black paper, I am going to use a light trap, so that there is no possibility of a shadow affecting the sharpness of the image edge. The idea is that a slant edge on a gradient gradually transitions across pixels on the sensor, as you move down the images beneath, the slight angle of the image means that the high contrast edge moves slightly relative to the grid of the pixels. The processing finds the gradient, rotates the image so that it is vertical, and then finds the average change in pixel value in the transition from left to right on the image. This is a digital equivalent to calculating the accutance. If the lens does not attenuate the sharpness of the original object, then the gradient would be completely flat until the transition where it would peak, and then it would be completely flat again. Practically, the gradient is kind of wave like. From the wave that is formed, a calculation is performed which works out what the lens can see in terms of sharpness...what frequencies have to be present in that wave, and how strong the frequencies are. The sharper the gradient, seen in the image, the greater the presence of the highest frequency waves...ie the higher the resolution capabilities of the lenses. Again, whats nice about this is that its not really distance related (a line looks like a line wherever you are standing, whereas a pair of lines doesn't), and thus does away with Edmund resolution targets. On the flip side, it involves a bunch of software and is less intuitive. I hope the explanation above is not too much like voodoo.
Its literally a simple question of generating images like these:
Pentax M 135 3.5 at 3.5
Pentax M 135 3.5 at 5.6
In the first image, its clear to see the colour fringing. The second image at f5.6 is sharper. Quick MTF said that for the 30% contrast level, the resolution was 1650LPH, in photozone.de speak...and this is with a dodgy test target....if I get a guillotine to ensure that the edge is dead straight, its possible this will be even higher. However, if you download the images and run QuickMTF now, you wont get the same results, as I have cropped the images. However, you will get .335 cycles per pixel...there are 172 pixels per mm, so this means 58lp/mm. This highlights the benefits of something like an Edmund test chart, in that you can see the bars with your eyes, and read off the resolution from a conversion chart. I guess this is work in progress. Aside from absolute lens resolution measurement, its useful just to go through the process comparing lenses to understand that the difference between good lenses is probably less than the variation in my ability to take a sharp shot, even when I am trying hard.
Software:
http://www.quickmtf.com/ http://mtfmapper.blogspot.co.uk/ http://sourceforge.net/projects/mtfcalculator/
MTF calculator is the quickest, but you can only use it for relative measurement. MTF mapper looks interesting but I have not had it working yet. Quick MTF is probably the most refined thing I have found, but its pretty expensive in comparison to other software...its quite tempting for me to actually write something myself. There is some shareware called imagej, and I might try taking the code into R.
PS, I am pretty sure my neighbours are wondering what the hell I am doing taking photos of black and white wedges!
Last edited by whojammyflip; 02-27-2014 at 02:46 PM.