In photography, the "holy trinity" is not a religious entity but rather three professional zoom lenses often favored by portrait, wedding, event, photojournalism and other photographers who need fast optics combined with high optical quality and robust construction. Typically, the three are: 70-200mm f2.8, 24-70mm f2.8, and a wide zoom, usually 16-35mm f2.8. However, there is another trinity, of the optically slower and lighter weight f4 variety, albeit still made with robust build quality. This trio is more often used by landscapers, hikers, travel photographers and others who favor lighter weight, more compact lenses that do not sacrifice high optical quality and durable construction. The compromise is that although somewhat similar focal ranges are covered, the maximum aperture tops out at f4.
One exception to the duplicated focal length range is the widely popular 24-105mm f4L zoom (24-120mm from Nikon). This lens sacrifices one f-stop in optical maximum aperture but offers a wider zoom range than its 24-70mm f2.8 cousin. For some uses, expecially "walk-around" situations, this can mean carrying one lens instead of two. Also, this zoom range allows one to add a 100-400mm f4.5-5.6 zoom for a highly useful 24-400mm zoom range in only two lenses. Unfortunately, none of the 24-105mm f4 offerings currrently on the market are up to the optical quality of the 24-70mm f2.8 alternatives, the later of which typically cost twice as much and may actually be heavier.
Sadly, it appears that lens makers have decided that there is not sufficient demand for an optic of the highest quality in this segment. Nevertheless, the 24-105 zoom remains popular, and it can deliver acceptable optical results for many uses, expecially when stopped down to f8 or f11. We routinely test every new lens acquisition for resolution and other parameters, and lately we've begun to acquire loaners of lenses we are considering. Thus it was that when Canon announced a new zoom in this mid range in their new RF mount, we decided to compare the new offering, via loaner, with our existing EF 24-104mm f4, the original version which was introduced in 2005 and which has been a staple for landscape, travel and everyday general purpose photography since. The hope was that the new RF mount version, introduced in 2018, would offer an upgrade in optical quality in addition to providing enhanced image stabilization and full compatibility with Canon's new R-series cameras, one of which, the recently introduced R5, has proved to be an outstanding all-around camera body.
Sadly, to quote Roger Cicala of LensRentals, "expectations proved to be a down payment on disappointment."
Our test setup and process was described in detail in Testing Camera Lenses, and a specific example of comparing two lenses is recounted in Testing the 400mm DO II. Thus, there's no need to repeat the discussion in detail. However a brief summary is in order. The test setup consists of camera, lens, tripod, multiple flashes controlled remotely, with modifiers (umbrella, softbox), and a USAF 1951 resolution chart. This chart has long since been superceded by the ISO 12233 chart, but those are pricey.and, surprisingly, do not lend themselves to comparative numerical answers. For our uses -- simple go/no go evaluation -- the older chart suits admirably.
Test image were exposures at f4, f5.6, f8, f11, f16 and f22, for each of the following focal lengths: 24mm, 35mm, 50mm, 70mm and 105mm -- a total of thirty test images for each lens. Image stabilization was turned off, and shutter release was on a 10 second time delay. Tests were performed using both mechanical shutter wth flash and electronic shutter, which does not allow flash use. The results reported below are with electronic shutter, a combo that should provide the best possible image quality, albeit at the sacrifice of some shutter speed -- of little importance given that a tripod was used, and on a concrete basement floor at that.
Lenses were tested and swapped out at each focal length before the tripod was moved and set up for the next focal length. This insured that, for each focal length tested, the camera and lens combination remained at precisely the same position and orientation to the test chart for each lens. One small boo-boo did occur; the EF lens was inadvertently tested at 28mm rather than 24mm. Once the discrepancy was discovered, the tripod had already been moved and the gear prepared for the next focal length. As a result, the EF lens results are likely a little worst than reported.
The USAF 1951 resolution test chart contains a total of individual 25 resolution test blocks, 21 of which are arrayed in an "X" pattern, with four added target blocks midway along each of the four edges. One can use the individual blocks to evaluate resolution at various positions around the image frame. Each block is further subdivided into groups of six small bars, paired three-by three at right angles and of diminishing size, The small groups are organized at two levels, with each upper level being composed of six smaller groups. This permits a determination of resolution performance based on the smallest lower level grouping that can just barely be discerened. The outer groups are labled as, starting with the largest, "-2", "-1", "0", "+1", +2", etc. The smaller or inner set of blocks are labeled as, for example, "1.1", "1.2", "1.3", "1.4", "1.5", and "1.6", where larger fractional numbers imply better resolution.
For 35mm full frame cameras, the outer group labeled "1" contains the smallest subgroups that can most often be just barely resolved. Tests such as this are dependent on the resolution of the camera sensor. Thus, when comparing lenses it is important to use the same camera (sensor) for testing each. The resolution in the center of the image circle for a very good lens tested on a twenty megapixel camera might resolve the second from largest line group within the six level "1" level set. In this case, the resolution would be assessed as "1.2". As we were to discover, moving up to the 45MP R5 as our test camera enabled resolution values of "1.6" in center-of-frame in some instances. Sad to report, for the two lenses tested, corner resolution in some worst case apertures could not be resolved in any of the "1" level subgroups, a rating recorded as "1.0". For the spreadsheets below, the initial "1." prefix is dropped and a resolution of, e.g., "1.4" is entered simply as "4".
It is possible, using a formula printed on the chart, to convert measurement to line pairs per millimeter of resolution. However, this value will inevitably depend not only on lens resolution but also sensor pixel density. Further, LPPM values are not needed in order to fulfill the objectives of this test -- i.e relative comparison of resolution performance as a basis for a go/no go purchase decision.
While it would be possible to analyze resolution data from each of the 25 test blocks, it was decided to concentrate on the center block (center of lens image circle), the four corners and the four edges. Although the final results might have varied a bit with the inclusion of more data, this selection of nine test areas gives a pretty good view of the central performance of the lens as well as of all corners and along each edge.
Resolution data was entered into a spreadsheet, initially arranged for direct comparison of each of the thirty test points for each target block: six apertures, five focal lengths and nine test chart sites -- a total of 270 data points for each lens. Unfortunately, there is a degree of subjectivity in assessing each resolution data point. For the sharper areas of the lenses, chart bars were usually crisp and distinct down to the point were no further resolution was possible. However, for worse performing parts of the image field, the bars were often fuzzy and far from crisp. In the end, the criterion used was, can all six test bars at a certain level be resolved, whether crisp or fuzzy. While this may not give a fully accurate assessment of a lens, it does provide a consistent and unbiased basis for evaluation. (And, there are sharpening algorithms that can extract information from those fuzzy bars.)
In order to compare the two lenses, a separate spreadsheet was construced by subtracting the RF lens score from the EF lens score for each of the 270 possible test conditions. In this second table, positive differences mean that the EF lens tested better and negative values imply a better RF lens resolution. A value of "0" implies that the resolution, good or bad, was the same for both EF and RF lenses. The final result was color coded for ease of quick visual interpretation.
As with any test of this nature, only one lens of each type was tested. Unfortunately, there is often significant sample variation across zoom lenses, so there is no implication that this test reflects all lenses of a particular type. Nevertheless, the surprising overall result is that my copy of the ancient EF 24-105mm f4 lens is superior to the RF sample tested. This is especially true at the edges of the frame and into the corners -- although there, neither lens performed up to the level of the best 24-70mm f2.8 lenses on the market. The newer RF lens is a bit better in the center of the frame, and it would take processing of more test targets to ascertain how far out toward the edges and corners this slight advantage extends.
However, the key point from my perspective is that for my uses in photographing scenics and architecture as well as general purpose subjects, usually stopped down to f8 or f11, the older EF version provides a flatter field and is therefore to be preferred. This surprising result has saved the cost of swapping out the older EF lens for the newer RF version -- and garnered a bit of good will with our chief financial officer, AKA wife!
Addendum: Lenses and Tests
A number of companies make long midrange zooms comparable in cost and performance to Canon's offerings. The list below includes those offered by the big three makers, Canon, Nikon and Sony.
Many of these lenses have been tested online, and some test sites even provide a comparison tool whereby two or more lenses can be compared side-by-side. A search of the following web sites will turn up useful test information on one or more of the above lenses.
|© 2020 Michael W. Masters Return to top|