The Holding Tank
Home Page Photography
Photography RSS 2.0 Feed
| Photos & Themes RSS | Hardware | Technical | Links | Books and Magazines | Downloads |
| Lens Data and Reviews | Lens Selection | Long Lens Selection | Wide Range Zooms | Consumer Zooms |
| Macro Equipment | Tilt and Shift Lenses | My Bags | My Old Hardware |
| My New Hardware | Lens MTF Comparison | Lens Tests | Weather Protection | Canon AF System | Camera Tests |
| EF 28mm f1.8 USM | EF 50mm f1.4 USM | EF 100mm f2.8 Macro USM |
| EF 180mm f3.5L Macro | EF 180mm f3.5L Macro + EF 1.4X II | EF 180mm f3.5L Macro + EF 2X II |
| EF 200mm f2.8L II | EF 200mm f2.8L II + EF 1.4X II | EF 200mm f2.8L II + EF 2X II | EF 200mm f2.8L II + EF 1.4X II+ EF 2X II |
| EF 300mm f4L IS | EF 300mm f4L IS + EF 1.4X II | EF 300mm f4L IS + EF 2X II | EF 300mm f4L IS + EF 1.4X II + EF 2X II |
| EF-S 18-55mm f3.5-5.6 | EF-S 10-22mm f3.5-4.5 USM | EF 17-40 f4L USM | EF 24-105mm f4L IS USM |
| Primes | 17mm Region | 24mm Region | 28mm Region | 35mm Region | 50mm Region | 100mm Region | 300mm Region | 400mm Region | 600mm Region |
| EF 180mm f3.5L Macro + TCs | EF 200mm f2.8L II + TCs | EF 300mm f4L IS USM + TCs | Tests With the Kenko TC | Macro Tests |
| Filter Effect on Sharpness and AF | Flare and Filters | Image Stabilizer Tests | TC Sharpness | Best of the Best | Method | Understanding the Results | Real World 100% Crops | Tripod Tests | L-Plate Test |
| EF 28mm f1.8 USM | EF 50mm f1.4 USM | EF 100mm f2.8 Macro USM |
| EF 200mm f2.8L I| USM | EF 300mm f4L IS USM | EF 300mm f4L IS USM +1.4xII Contrast |
| EF-S 10-22mm f3.5-4.5 USM | EF 17-40 f4L USM | EF 24-105mm f4L IS USM | Brand/Grade Test | Filter Data | Protective Filter FAQ |
Test Setup Validation
This section looks at the resistance of 3 zoom and 3 prime lenses to flare and ghosting under controlled conditions, with and without UV protective filters. For the effect of filters on AF and sharpness see this experiment.
The test conditions simulate fairly harsh conditions with a strong light source in the frame, the frame being largely underexposed. This is expected to be representative of outdoor photography at night and is probably for challenging than sunrise and sunset shots in terms of dynamic range.
Manufactures Filter Data is summarized here.
Investigation of anti-reflection coating theory here.
A test target is used to provide something to focus on with a largely uniform brightness. An incandescent lamp with a 60W bulb is used as the strong light source. The lamp has a directional hood and is placed pointing at the lens and causes minimal illumination of the test target surface.
The following readings are taken using a Gossen light meter at 100 ISO:
White Surface of Target: ~6EV
Light Source at Hood Opening: >>18EV (off scale)
~15LV (incident) [Estimated 20EV [Note 1]]
So there is something like 14 stops between the bright light flux density and that of the general scene. The exposure is set so the white area of the test target is underexposed by around 1 stop to help bring out flare near the light source.
As far as possible the lenses are tested at a constant magnification of something less than 1:27. The camera (20D) is placed on a tripod and focused on the test target at each tested focal length, wide mid and tele for zooms. Each principle stop value from wide open to f16 is exposed for in manual mode. The filter is then removed/replaced without disturbing the setup and the exposures repeated. RAW capture is used.
Note1: The estimate of 5 stops between incident and reflected readings is based on an incident and "reflected" reading of the sun's illumination.
Taking a crop of the light at 14 stops below the test exposure shows the filament of the light bulb is the brightest light source, about 2 stops above the exposure level, i.e. 16 stops above the test exposure.
This makes the filament about the same angular size as the disk of the sun for a 10mm lens, although a rather odd shape.
In the case of the disk of the sun being in frame (or at least not shielded by the lens hood) the disk of the sun has a luminance of about 1,000,000,000 cd/m2 at zenith at the equator and that of a sunlit scene is around 100,000 cd/m2 [Ref]. This is a source to scene luminance ratio of 10000:1 or about 13.3 stops. Comparing the test setup to this situation a slightly pessimistic brightness difference (~14-16 stops) has been generated, so it seems this is a representative test situation but at least under controlled conditions without having to consider changing weather or the movement of the sun with time. The core of the test bright source will be rather large for lenses longer than 10mm on APS-C or 16mm on full frame.
Below is an example shot at f16 with the EF-S 10-22mm @ 10mm with the disk of the sun in frame. The exposure was set for the sunlit grass.
Here we can see some green flare opposite the sun's disk on the lower right.
A similar geometry can be seen in test position 2 under at the same lens and exposure conditions showing similar flare..
So it is concluded that the test setup is representative of a scene lit by a bright light in frame such as the disk of the sun.
It is worth noting that the green response is due to the lens and not the UV filter fitted in both cases. The glare of the sun in fact hides the white responses around the light source that are due to the filter.
Digitally reducing the real world shot's exposure by 4 stops reveals some responses in the glare of the sun which might be due to the filter.
The RAW files are renamed and the white balance set to a value of 2400K +5 tint for all exposures to compensate for the colour temperature of the domestic incandescent lights. The exposures are then resized for the web with some highpass sharpening to compensate for the downsize operation.
For the lens wide open and f16 exposures an additional step is performed to provide a difference between the no filter and with filter cases. This is done using Photoshop CS2 and layer blend with difference. To enhance the visibility of differences all files were adjusted with the Brightness/Contrast menu by 50 and 60 respectively.
The resulting files are also down-sampled for the web in the same way. Small differences in the shot may result is some edges being visible in the difference pictures.
An example of a no difference case is shown below. Here the same image is differenced with itself with a 1 pixel alignment error in both the vertical and horizontal directions prior to down-sampling.
The results can be explored using the menu items above. In the case of zoom lenses a secondary focal length menu will be found on the results page.
As might be expected under such harsh lighting conditions all the lenses flare to some extent, the effect tending to become enhanced most at f16.
In most cases the addition of a filter made little or no difference, in some cases a difference was observable although this was a second-order issue, the lens itself producing most of the flare.
Some observations are collected below but the reader is encouraged to look at the results themselves.
Notable problem areas:
EF 28mm f1.8 USM:
There is a faint green general flare in the opposite quadrant from the light source narrowing and weakening as the lens stops down; gone by f11. This seems entirely due to the lens and the filter does not seem to have any adverse effect. General contrast level across the frame are unaffected by the filter.
EF 50mm f1.4 USM:
There is a blue-green response due to the filter from wide open but becoming diffuse and more weak from f4. This response remains local and has no clear effect on general contrast levels in other parts of the frame due to the filter. This kind of filter response seems to occur with lenses that have well recessed front elements. There do not seem to be any other responses from either the lens of the filter.
EF 100mm f2.8 Macro USM: There is a small amount of ghosting in the opposite corner of the frame caused by adding a filter at f2.8, this has vanished by f5.6. However this ghosting is of a similar order to the lens native flare and ghosting.
EF 200mm f2.8L II: Flare is well controlled and there is no visible ghosting with or without a filter. The difference image suggests there is a small difference in the diffuse flare around the light source at f2.8 between the filter and no filter case, but this is not noticeable by eye in the basic shots.
EF 300mm f4L IS: Technically this lens probably does not need a filter for protection and the front element is probably a factory fitted protective element. The addition of the filter causes two areas of intense ghosting at all f-stops. Even in this case, if the light source was 3-4 stops less strong the issue would be invisible. The strong ghosting observed is probably due to the relatively recessed front element on this lens.
EF-S 10-22mm at 10mm: By f16 lots of additional reflections of the lens flare can be seen in the with filter case however the general image contrast is not affected. The problem is reduced considerably at f5.6 and wider. If the light source was about 3 stops less bright the filter contribution would be insignificant. The same issues can be seen at other focal lengths but are less strong.
The tests for this lens also include 5 different light source positions at 10mm f16. In some positions a green lens response or in the centre a halo is revealed. However these all seem to be due to the lens and not the filter.
EF 17-40 f4L at 17mm: Very similar to the 10-22mm at 10mm but not quite so marked. There was negligible effect at 28mm and 40mm by adding the filter.
EF 24-105mm f4L IS: The flare is much worse than the other tested lenses. This is consistent with other test results for this lens but is not as bad as the flare issue Canon issues a recall for [ref 1] [ref 2]. The lens flare response is large enough there was no obvious degradation by adding a filter.
In general addition flare and ghosting due to a good quality UV filter is of the same order as the lens or better, they will however tend to pollute additional parts of the image in addition to artifacts arising just due to the lens. Effects from both sources tend to be larger at smaller apertures. However there are some exceptions to both of these. The artifacts tend to be localized reflections with no measurable impact on overall contrast outside the local area of effect.
As some of the filter produced artifacts are clearly the result of reflections between the lens front element and the filter, it is possible that a lens with poor front element coatings producing significant reflections to the filter may fair worse with a filter fitted than a good quality lens.
Also individual equipment may vary and the effect of a strong light source has only been investigated in the corner of the frame other than for the 10-22mm at 10mm f16 where five positions were tested.
There is some evidence that lenses with well recessed front elements may be more prone to detrimental effects of filters than those with front elements close to the filter glass.
To see why, take a look at the diagram below:
We can see in the case on the left with a well recessed front element, any light reflected off the front element and then reflected again off the filter arrives back to enter the front element some distance from the original ray. On the other hand, with the filter close to the lens the second order reflection enters the front element much closer to the original ray.
The effect of this is in the case on the left the ghosting image can be clearly separated from the light source in the image, but for the case on the right the light source hides any ghosting image which is effectively coincident with the source.
The test was conducted with about 14-16 stops difference between the strong light source and the general scene. It was noted that any filter responses (and most lens responses) would be insignificant if the light source was 3-4 stops less bright. The test is thought to be representative of strong lighting situations such as the disk of the sun being in the frame.
From this the conclusion is that for normal daylight usage a good quality filter should not present a problem except if the sun is in the frame with no attenuation from clouds. Even then the lens flare may be worse than the filter's. In the case of sunsets and sunrises, with the sun in the frame, some small artifacts might just be discernable in shadow areas. Even in these cases the lens may be just as bad as the additional effects due to the filter.
Where more care is probably needed is in night photography, particularly of street scenes or performing arts photography where strong spot lights might be directed at the audience. Even in these cases the lens flare may be worse than any additional flare from a good filter. Some tests are advisable before commencing a significant body of work under such conditions using a filter. It is also worth trying to not use too small an aperture to minimize any ghosting from either source.
What the optimum strategy with respect to using UV filters as a protective element depends on what then photographer is shooting.
If he is working largely without a strong light source in the frame, and assuming the lens hood is effective then a good quality filter is unlikely to have any detrimental effect even if the subject is heavily back lit; he may thus elect to fit protective filters and only remove them if including a strong light source.
Someone habitually including strong light sources in the frame may find it prudent to not fit a filter unless he finds himself working in a hostile environment.
My personal situation is the former where the value of a protective element fitted by default covers 99% of my photography.
Finally it should be noted that any lens is likely to flare if the luminance difference between a light source and the scene is large enough. A real world example.
Last Updated 27/06/2008
All Content © 2005-13 Lester Wareham All Rights
All material is supplied as is and without warranty,
use at your own risk.
All opinions stated are the authors own.
All quoted information remains the copyright of the
Note this email gets spammed a lot so if you want to get your message through please prefix "Holding Tank" in the title