A Bit of a Grumble Frustration with RF Lens Designs
I have a bit of grumble with the way Canon’s doing things with their new RF lenses right now, and it boils down to a couple of overreaching points.
First, at least in the lower tier lenses, Canon is now strongly using distortion correction to, well, make things work. Almost all of the lower end zooms simply don’t cover the full frame sensor area when at their widest focal lengths. This is the case for both of the consumer level RF zooms; the RF 24-105mm f/4-7.1 IS STM, and the RF 24-240mm f/4-6.3 IS USM. At 24mm both of these lenses mechanically vignette horrifically, and rely on either in camera or post processing distortion correction to generate a full frame image.
Now all of this isn’t to say that I have a problem with distortion correction. I use it all the time on my L lenses, to correct distortions. But, for me at least, correcting “distortion” that is an image that doesn’t even cover the full sensor area is a bit of a stretch (pun intended).
That said, at the same time, I can kind of understand the situation. Physics, and covering useful zoom ranges, mean that there’s limits to just how small you can make a full frame lens.
Simultaneously, there’s clearly a lot of demand in the market for smaller and lighter lenses, especially at the lower tiers. My mom is a prime example of this, having moved from crop DLSRs to an EOS M mirrorless camera in no small part because it, and it’s lenses, are smaller and lighter.
And when push comes to shove, compromises have to be made.
So entry level lenses get slower than traditional apertures, f/4 to f/6.3 or 7.1 instead of f/3.5 to f/5.6, and they get mechanically vignetting image circles. And if you don’t want that kind of behavior, there’s always the L lenses.
What really struck me as annoying though is the test results that Brian over at The Digital Picture has seen with the RF 100mm f/2.8L Macro IS USM. Well there’s actually two points that come up with this lens.
First, is his complain that the lens focuses in little jumps instead of smooth infinite steps. Of course, this is most noticeable at the highest magnifications.
This behavior is simply a product of linear focusing nano-USM systems and focus by wire. Compared to the older helical systems, linear focusing motors have more room to be less precise. What he’s seeing is a combination of limited focus step sizes, and keep in mind that at 1.4x magnification we’re talking about 10s-of-micrometer scale movements of the focusing system to make millimeter scale movements in focus.
The bigger problem, for me at least, is the lens’s focus shift and Canon’s response to it.
Now again, I understand engineering well enough to understand that there’s always compromises in things. Lenses are not aberration free, and it’s probably not a good idea to expect complex lenses pushing into new territory to not exhibit some of those aberrations. In this case, it looks like residual spherical aberrations cause the focus to shift as the aperture is stopped down.
So far as I’m concerned, focus shifts are a problem; full stop. They’re doubly a problem when shooting at high magnifications where the shift may simply be significant enough to not be covered by the lens’s depth of field. And that seems to be the case with the RF 100mm macro.
When Brian noticed this, he got in contact with Canon USA and ultimately Canon Japan to determine if the problem was specific to his lens, or if it was specific to the design. And the response that get got back from Canon was that it was not a flaw in the specific lens, and that it was due to the high 1.4x maximum magnification. Moreover, apparently there was no indication of corrective action being taken.
And this is where I find I start really getting frustrated with the whole situation. Cameras and lenses today aren’t the simple dumb lumps of glass and metal that they were in years past. There’s full on microcontrollers in every lens now, with enough complexity that they even get firmware updates from time to time.
Focus shifts due to residual spherical aberrations aren’t uncharacterizable behavior. Canon could certainly map out the focus shifts as a function of focus position and aperture value, and add in a table or corrective function to the lens’s micro controller to fix the situation. In fact, I suspect though I haven’t dug into it enough to be sure, that the focus shifts could be modeled with a linear equation that’s simple enough that any micro controller should be able to do handle it.
In short, a firmware update could easily fix the problem by adjusting the focus when the aperture is stopped down. The lens knows precisely where the focus position is, and it know what aperture it’s commanded to stop down to, so it’s not like it should need any external assistance.
My frustration is compounded, thanks to the “focus by wire” design of the lens — though this would apply to basically all other mirrorless lens on the market now too. On my EOS R5, when the DoF preview button is depressed, the camera will not allow focus adjustments to be made to the lens; either AF or MF. And since, everything goes through the camera, there’s no way around this.
With mechanically coupled EF lenses, you could always turn the focusing ring, and always move the focusing groups in the lens. That’s simply not the case with a focus by wire design. At least not unless the camera is designed to allow it.
This effectively renders the best non-guess workaround of focusing slightly in front of the subject, moot.
Compounding matters, the RF platform’s dual pixel AF sensors are more than capable of focusing at apertures much slower than f/2.8. And while focusing at the widest aperture has significant benefits in insuring more accurate focusing, in this case, that harms the end result when the focus shifts.
In short, if Canon couldn’t fix the issue by applying a corrective function in the lens’s internal firmware, there’s a number of ways the camera’s could be updated to workaround the issue instead. From enabling focus adjustments to be made when the DOF preview is engaged, to having an option to verify or refocus certain lenses, or certain lenses at certain object distances, after stopping them down.
In a strange way, the really frustrating part about all of this is that on one hand, you have the lower tire lenses being designed from the ground up to be reliant on software corrections to preform properly. But then software corrections aren’t apparently an answer to insuring that higher end lenses preform maximally well.
P.s. The thought did occur to me that part of the problem here is sensor resolutions outpacing the design requirements for the lenses. According to the Digital Picture’s Review, on a 20 MP EOS R6, the focus shift is noticeable but not as significant, and with sharpening results in reasonable images.
Lower resolution cameras get away with this because the pixel density allows for larger blur spots to remain in the pixel boundaries. As a result, a lens can be softer, or the focus can be less precise, and the image still appears pixel sharp.
This isn’t that far off from the situation that many photographers have found themselves in moving from low resolution (<=20 MP) cameras, to high resolution (>30MP) cameras, and finding that lenses they thought were absolutely tack sharp simply weren’t that good.
There’s still quite a number of places where old, film level, calculations are still being used. Depth of field, for example, still uses assumptions about enlargement ratios and print viewing distances that were developed with film. And, for the most part, at least in the case of depth of field, the assumptions still mostly hold, as the end result is still a print that’s viewed form similar distances, by the same eyes that viewed film enlargements 50+ years ago.
And in the end, the question has to be asked; what were the design tolerances that Canon’s engineers were aiming for?
Great post! I discovered focus shift on my RF 35mm 1.8 STM with an R5 when focused around 2m (i.e. problem is more pronounced when focused close) and I stopped down to f4 – I was surprised to see the image soften!? I experimented with shooting a flat ruler so I could see the direction of the focal plane shift and the issue was consistently reproducible. The RF 50mm 1.8 STM has the same issue, whilst the RF 24-105 4.0 L does not – I attribute this to the smaller 4.0 aperture as a starting point, thus stopping down to f8 will more likely have the focus shift effect cancelled out by the increase in depth of field at f8 (the RF 35mm 1.8, which was the worst at f4, had this same cancelling out effect by f8). The RF 24-105 4L may also have been engineered to reduce this phenomenon, given its higher price point compared to the STM lenses. After some web research it became apparent that whilst this is a very real issue, many had not experienced this which I’ve concluded is down to some users misunderstanding the conditions to recreate, with many others blissfully unaware, i.e. with a lack of Canon users kicking up a stink, Canon have no incentive to fix this.