I write a lot about the fact that every copy of a lens is slightly different than every other copy. Now don't get me wrong: I'm talking slight differences that are barely detectable, not good or bad.
Even when things appear to be in the "bad lens" area, they usually aren't. The majority of the time when a person thinks something is wrong with their lens, nothing really is.
Expectations are out of line or the learning curve of a new lens hasn't been mastered yet. Even when there are real problems, usually they are simply autofocus problems in which the camera-lens combination just requires a bit of autofocus microadjustment. But every once in a while something is amiss with the optics of a lens.
I tend to write about "bad lenses" a lot because my major job here at Lensrentals is to check our stock, make sure the lenses are optically as they should be, and fix them if they aren't.
In one post, I mentioned that it wasn't unheard of for lenses to come back from repair with optical issues. That led to a lot of emails from people who didn't understand how lenses could be less than perfect. My verbal explanations didn't seem to help explain what was going on, so I thought perhaps showing you might clarify things.
Before we start though, the majority of this post is a Geek Level 4 article.
If you aren't into the mechanics of how a lens works and what might cause problems with it, you will find the section on "Optically Adjusting the Lens" incredibly boring. You might want to just skim over that section and then read the "What's the Point" section at the end, which covers briefly why I think this stuff is important to anyone who owns a lens. If you don't own a lens, then you're definitely in the wrong place and need to hit the back button on your browser a couple of times.
I want to be clear that this is NOT a how-to-optically-adjust-your-own-lens article. First of all, what we've learned, we've learned by reverse engineering.
We don't have Canon's, Nikon's, or any other company's in-house software and hardware that allows them to read lenses and predict necessary adjustments. We don't even have the in-house manuals on how they actually do it. We've found methods that work for us, but that doesn't mean it's the best way or the right way. I'm certain it's not the fastest way. It's just the way we've figured out how to do it.
Even our "shade tree" methods require a lot of specialized equipment. We do our adjustments using repeated measurements with a $15K lens-test projector and a $12K Imatest optical testing setup. Even that isn't ideal, and we have optical bench equipment ordered to augment our abilities.
There is simply no way to do this kind of stuff at home with some test charts or photographing a brick wall. And finally, every single lens has different places and methods for optical adjustments. The lens we're doing here is one of the simpler ones.
Identifying a Bad Copy
I'm going to use a Canon 24-70mm f/2.8L lens as an example, mostly because it's a lens we know well and partly because adjusting it is pretty straightforward and the adjustments are easy to photograph. Not all lenses are as easy to work on. The copy we're using for this demonstration has a fairly typical story: it was dropped, causing the filter ring to bend.
The ring is easy to replace but from long experience we know that dropped lenses often have problems that go well beyond the dent noticed on the outside, so all dropped lenses get optically tested before they're put back in stock. This particular lens looked fine at 24mm but showed the pattern below when run through Imatest at 70mm.
From our routine testing, we know that a Canon 24-70mm f/2.8L should resolve around 700 line pairs / image height in the center and 560 lp / ih weighted mean, and this lens does neither. It also is very soft in the corners with several corners under 200 (250 is the absolute minimum). When we photographed a Siemens' star chart with it, we saw a typical decentered pattern.
Optical Adjustments for This Lens
The Canon 24-70mm f/2.8L had a number of adjustments that can be made very easily. Several involve the front element, which is probably why it tends to get decentered with a fall or drop (like a lot of lenses).
With the filter ring removed, it's apparent that the front element is a centering element: loosening the three front screws allows us to recenter the front element by sliding the element under the screws in whatever direction is needed.
Looking at the side of the front element, there are three screws that, when loosened, allow the element to rotate along a ramp, moving it slightly forward or backward from the rest of the lens.
Another set of three screws on the front are surrounded by oblate (thicker on one side than the other) collars. Adjusting the collars tilts the lens forward or backward at that location.
There is a similar set of collars located under the focusing ring that tilts the second lens group.
Performing the Adjustments
These adjustments are all rather precise. A movement of 0.5mm when centering the front element can make a dramatic difference, as can a turn of 30 degrees of an oblate collar or a movement of 1mm on the sliding ramp.
Since the lens seemed decentered, we started by loosening the three front screws and moving the element in the direction of the flare seen on the Siemens' chart. This a fairly tedious process (since we don't have an optical bench yet) that involves moving the element, remounting the lens to a camera to reinspect the image, and repeating until we seem to have good centering.
Once centering appears good as evaluated on the Zeiss-Siemens' star chart, we retested the lens with Imatest and got the following results.
The centering has improved our corners somewhat (although not to acceptable levels) and actually lowered our overall resolution.
The charts I'm printing here only are showing vertical resolution at 13 samples, but in reality we're checking vertical and horizontal resolution at 33 locations. (I can't reproduce those on images of this size - it just becomes a massive jumble of numbers.)
We also saw that vertical and horizontal numbers were much more equal after centering, which confirmed we were on the right track.
With the lens centered, overall resolution was still poor at 70mm, although good at 24mm (I'm not showing the 24mm results---trying to keep this post reasonably brief). From experience with a lot of dropped lenses, we felt adjusting the sliding ramp would be the best way to improve 70mm resolution. This can be done fairly quickly on the lens-test projector since we can evaluate sharpness in real time as the front element is adjusted.
We then repeated Imatest, which showed acceptable resolution numbers except for that there was still some obvious tilt, affecting the right side of the lens and especially the lower right corner.
We went back to the lens-test projector and adjusted the tilting collars on the front element, repeated the Imatest, and went back and adjusted some more. Eventually, we got everything exactly where we wanted it, as the image below shows.
Of course, we had to double-check the results at other focal lengths, to make certain the adjustments we'd made hadn't done something negative at shorter focal lengths.
Since we had not needed to make adjustments to the second element we didn't think there would be any problems at other focal lengths, and there were not. It's not surprising that dropping the lens on the filter ring affected alignment of the front element, but not the second element.
To give some perspective, though, I've shown you only four of the actual 16 Imatest shots taken.
The entire adjustment from start to finish took about 90 minutes (including disassembly and reassembly), which makes this a fairly quick, straightforward adjustment.
With more complex lenses, it's not unusual to spend three or four hours returning a lens to optical perfection---some lenses require major disassembly to make an adjustment, then partial reassembly to test that one tweak.
There are several, actually. For some perspective though, let me assure you this lens is one of the simplest and most straightforward to adjust. Other lenses have up to a dozen (that we know of) adjustments that can be made.
In addition to the straightforward adjustments shown here, there are also multiple shims to adjust spacing between elements or between the lens and the camera (for proper infinity focus). In some cases, certain parts in the lens (like the rear mount) come in a variety of thicknesses, acting in the same way that shims would.
My first point is that given all of the adjustments that can be made with each lens, it is inevitable that each will be slightly different, even when "tuned" to its best resolution. That difference, though, is too small to see in a real photograph. It's only apparent with a ridiculous degree of pixel peeping or using overly accurate test equipment.
If you notice in our last graph, above the right lower corner is still slightly softer than the left upper corner. Not only is that as good as we could make this lens, but it's also far better than average. As I mentioned above, the average Canon 24-70mm f/2.8 resolves 700 / 560 with a minimum resolution of 250 lp/ih. This one now is 733 / 650 with a minimum of 363 lp/ih.
Could you tell the difference between the upper left and lower right corner or between this lens and an "average" 24-70mm?
Well, obviously if you have Imatest or an optical bench you can. Maybe, just maybe, if you have high quality optical test charts, a system to guarantee exact right angles to the chart, and you pixel peep at 100 percent you could tell the difference.
In a photograph? I doubt it seriously. On the other hand, you could tell the difference between this lens before and after adjustment at a glance.
The second point is that making these optical adjustments is time consuming and complicated. For us, the time and effort is financially worth it: if we don't get the lens right, it can't be rented and we're going to part out an expensive lens. So spending three or four hours getting it right is worthwhile - we are going to loose over $1,000 if we don't fix it.
If we were a repair center getting a flat $180( or whatever the fee is to fix a lens), spending three or four hours to optically adjust an optically complex lens it would make us go bankrupt pretty quickly. Factory service centers, given their equipment and training should be much better at this than we are -- but unfortunately they aren't.
So far this year, we've had over 30 lenses that failed optical testing when they came back from factory service and went straight back to the factory for them to try it again. Over a dozen went back to factory service twice, and in most cases we ended up doing the optical adjustments ourselves. In several we just ended up parting the lens out because they couldn't be fixed (by the service center or by us). Keep this in perspective, though. That's out of several hundred lens repairs on thousands of lenses. It happens, but it's not common.
My third and final point is one that people do not like to hear. If you drop a lens hard enough to dent its filter ring (or the filter's ring if you have a filter mounted) check it carefully even though it seems to work fine.
Certain lenses are likely to have something jarred loose inside affecting its image quality after a drop. Often, like this lens, adjustments can be made to correct it. At other times though, the drop has caused a slight bend in a barrel or other piece that must be replaced to restore optical integrity.
People often ask me what they should look at when buying a used lens, and dents or other signs of being dropped are very high on my list.
Roger Cicala and Aaron Closz
Addendum: OK, I have to admit I was busted in an email which said "Nice article, but I'm calling BS on the why you do it part: you mostly do it because you're a gear-head and this is fun for you. We all know this." I am. It is.