MTF, or Modulation Transfer Function, is a scientific way of measuring lens performance, But isn’t that just stuff for Geeks? Well, no, it does have a bearing on a lens’ ability to produce sharp images that goes beyond the photographers technique and goes a long way to explain the difference in prices between apparently similar lenses.
Lenses performance can be measured in a number of different ways. An optical test bench can cost many thousands of pounds/euros/dollars and will give an accurate reading on any lens put through it’s system.
Computers can work out what a lens should be like from the design parameters. That is the way that Canon produce their MTF charts. This method gives a theoretical answer that does not take into account manufacturing anomalies.
Another way is to put an example of the lens in question onto a camera, shoot some targets that can be measured, and get the results from that. That is the way that we do it here at SigmaUser. It has some obvious advantages, in that you get to see how a real lens works in real circumstances, but it also has some disadvantages. The disadvantages have come about by the modern media we now use, that of digital capture of our images.
Why, you might ask, is digital a disadvantage? Generally speaking, it is not and that is why it has become so popular in such a short space of time. With lens measurements though, the sensor needs to stay the same otherwise the results vary, not through the fault of the lens but through the use of different sensors! The days when we could put a roll of the same film through a bunch of different cameras and get comparable results are now, sadly, long gone.
What this method does do however, is to highlight, at a glance, where and how the lens in question works best. By taking the results from each aperture (and the tests are done at each full aperture for simplicity) and showing them in a simple line graph it can be seen that each aperture produces a different result and the highest point of the graph is where the lens performs best.
The actual numbers down the side of the graph are not really that important, unless you are comparing one lens against another on the same camera. The numbers can typically be in LW/PH (Line Widths per Picture height) or Cyl/pxl (Cycles per pixel) and these can be measured at different frequencies with the higher the frequency the finer the detail. An example figure might be 1465 lw/ph at MTF 30. At a frequency of MTF 10 the number would be higher whereas at MTF 50 the number would be lower.
The same goes for different sensors where higher pixel counts will measure the lens and give different figures (normally lower numbers) than a sensor with less pixels. Therefore, the same lens, tested on say, three different cameras like a D40(6mp), D80(10mp) and D2X(12mp), will give three different sets of numbers, but, put into the graph, will give the same shaped curve.
We take two sets of readings, one near the centre of the frame and another towards the edge, and these two sets of figures give us the two lines on the graph. The third line, where shown, is the average of the first two.
So what are we looking for in the graph? Well, the first thing is the peak. That is where the lens is going to give the best results, at that aperture. The aperture scale goes along the bottom of the graph, so find the highest point and run your eye down to the scale at the bottom and read off the aperture.
The next thing the graph will show us is how good the lens is towards the edges, compared to the centre. A graph where the two lines stay almost parallel to each other without a big gap between them means that a lens is performing almost as well at the edges as it is in the centre. Ocasionally, a lens will show that the edge out-performs the centre at some points (this occurs most often with lenses designed for full frame use) and this is a good sign.
The ideal graph would show two lines, close together and almost flat. This, of course, doesn’t happen due to a thing called reciprocity coming in around the f/11-f/16 mark where the size of the aperture gets so small it causes a degrading of the image quality.
The fact that the numbers change with the type of sensor used, the aggressiveness of the anti-aliasing filter, pixel count etc. etc. does mean that you have to take the testers word for it as to wether the lens is good, bad or indifferent, the information it does give will show you how to get the best results from that lens.
Other areas of a lens test are more quantifiable. For instance distortion is given as a percentage with a + or – number. Positive numbers (+) mean the lens at that point shows pincushion distortion while negative numbers(-) mean the distortion is barrel shaped. Numbers above/below 1.5% will not be visible to the naked eye under most circumstances but if the numbers are bigger than 1.5% then the distortion will start to show.
Similarly, Chromatic aberrations are measured in pixels and, regardless of the number of pixels on a sensor, if the ‘fringe’ that the anomaly causes is greater than 1.5 pixels, it will start to show to the human eye when viewed at 100% on a monitor. This doesn’t mean that it will show in a print, which is more dependent on the accuracy of the printer. It is the concept of being able to view everything at 100% on a monitor that has highlighted the inadequacies of some older lenses.