dante stella stories photographs technical guestbook
|Mutatis mutandis convers a lot of territory, baby!|
|Can Photodo.com show differences in lenses?|
Photodo.com provides some unique information. When you look at a manufacturer's MTF charts for a lens, you will often see what looks like astronomical MTF numbers. Some of Tamron's zoom SLR lenses purport to have 95% MTF at 10lp/mm. Sure. Cosina is also good at publishing high MTF numbers. The ugly truth is that the published numbers are computed MTF, which does not take into account manufacturing tolerances. You can actually get a lot worse contrast from a lens than the manufacturer's MTF would indicate. The sole exception is Zeiss, which publishes measured figures.
Enter Photodo.com, an arm of Hasselblad which actually does MTF testing on consumer grade lenses. Photodo publishes MTF graphs which show you what you would otherwise need about $700 of testing (per lens) to discover. Photodo publishes two graphs for each lens, showing 10lp, 20lp and 40lp/mm. These are very interesting, because they show what the lens actually resolves (most times the measured MTF is below the actual; inexplicably, Konica Hexar RF lenses have significantly better MTF in testing than their own manufacturer claims...)
Photodo also publishes ratings, which are weighted averages of the MTF graphs. The trouble is that some people seem to believe that photodo's numeric rating is a determinant of how good a particular lens is. As set forth below, photodo makes its own assumptions about what is important and what is not, and the weighted averages it kicks out as its "ratings" are not necessarily meaningful.
Photodo.com comes up with some interesting results. Take, for example, a look at the 50/1.8 AF Nikkor ($100, rated 4.4 by Photodo) and the 50/2 Summicron-M ($900, 4.6). If you just looked at the weighted numbers, you would guess that the Leica is only 4% better, while costing 800% more. But if you look at the wider apertures, you will notice that the Summicron blows away the Nikkor. But due to the ways that Photodo weights the results, the Nikkor comes out much better off. Photodo bases its lp/mm testing on the following assumption:
This is something of an oversimplification with regard to the importance of the 10lp/mm level, since that is relevant at all sizes of print viewed from a normal distance. 40lp/mm is important if you are grain-sniffing a large print from 6 inches away, and Photodo acts as if this is the way people normally look at prints. Suffice to say, it isn't. To view any full print you have to back up a distance which increases proportionately to print size.
Here is what Photodo says about weighting:
When you read through photodo's documentation on its tests and weighting, it becomes apparent that photo assumes that every lens will be used in the same way with same mix of apertures. Another basic assumption is that all lenses are optimized the same way, and peak at similar apertures. To see how this breaks down in practice, we will look at a medium rangefinder telephoto, whose primary purpose is portraiture.
The exception which disproves the rule. While it may be true that many 35mm lenses (i.e. SLR lenses, especially wideangles) have peak MTF at f/8, and that many have poor MTF wide-open, this is by no means true of rangefinder lenses. Rangefinder lenses often have great MTF wide open and peak well before f/8. Let's take three 90mm f/2.8 lenses as an example.
Observe above (L-R) are the Photodo charts for the Leica, Contax and Konica 90mm f/2.8 lenses. You will notice that at the 10lp/mm level, there is no significant change in contrast with any of the three. They behave similarly at f/2.8 and f/8. They all hit 90% and above, meaning they lose about a 1/10 of a stop of contrast as compared to the scene without lens.
All three of these lenses are portrait lenses. Assume that in the real world, you are shooting portraits wide-open with any one of these lenses. Bear in mind that on prints, only the 10lp/mm number is relevant, since that determines acutance on a print viewed from a normal distance.
The respective weighted MTFs (across all apertures) at 10lp/mm (here the comparison is on output) are 92, 92 and 89, respectively. This is a 3% contrast difference, a difference that can easily be sucked up in (a) focusing errors (especially with the Contax G) or (b) shutter speed errors. Remember, Photodo tests are measured MTF (which elminates the camera body as a source of error).
Cross-section this another way - at f/2.8 (here the comparison is based on composition) across all resolution levels. Here, the Photodo testing shows 81, 81, and 76, respectively, a 5% difference and still within the province of focusing or exposure error (consider that one stop of exposure is 100%). Given the maximum accuracy of an autofocus system, it is entirely possible that the Contax could do worst in the real world. Or given the shutter speed accuracy of a Leica, if that body runs slow, a combination involving a Leica M body could be bottom of the barrel with either the Elmarit-M or the M-Hexanon.
But when you look at the Photodo final numbers, you can see that the weights it assigns to the various combinations (and here is where their mix of anticipated shots will vary), you would think that there would be a 21% spread in performance between the Leica (4.6) and the Konica (3.8), with the Contax falling in between (4.4). You would never any difference like that (or probably any difference) in the real world if you are shooting wide-open or making prints, and your weighted numbers might look more like 4.6, 4.6, 4.4. That assumes, of course, that the Contax can focus as accurately as a Leica M user with that lens at a meter.
What's the moral of this story? Photodo provides a lot of useful information in its measured MTF graphs, but the value of its weighted scores may be little. What you get out of the site depends on your ability to read the graphs and interpret what they mean to your shooting.