For this test a high contrast target is made from a light box and a small slide box lined with black felt. The test setup is photographed with no filter on the lens, a Hoya S-HMC UV filter (a multicoated filter) and a Tiffen Haze 2A filter (an uncoated filter as far as I can tell from Tiffen marketing material). A 70-200/4L lens is used since it is a good constrast lens that has good tolerance for flare. Most importantly I own this lens and happen to have 67mm filters of both types available. For each condition three shots were taken to verify there was no shot to shot variance. Only one of each is shown here.
So here we have the 70mm/4.0 shot with no filter. The average RGB value of the center of the dark box is 53,51,46. This is the base contrast level for this lens at this exposure. For those who care these were all shot in sRGB with daylight WB and parameter 2 set on the XT (that is, contrast, saturation and sharpening at 0).
This is with the Hoya S-HMC UV filter, the RGB is 54,52,47. This is a negligibly lighter tone than without the filter, but for what it is worth we can say we lost a smidgen of contrast. One would be hard pressed to see the difference.
This is the Tiffen Haze 2A filter. Let no one ever say coatings don't matter. Yikes, there is a clearly visible difference here. The RGB value is 96,93,80. This shows an uncoated filter causes significant loss of contrast compared to no filter or a multicoated filter.
Just for variety the same test was done at 200/4.0. Here, with no filter, the average RGB value of the center of the dark box is 41,41,37. Again this is the base contrast of the lens for this target. This seems to indicate the lens is a bit more contrasty at the long end, but perhaps this is also because the lightbox doesn't quite fill the frame as much as in the 70mm shots.
With the Hoya S-HMC we get 46,44,39. Now we see a more measurable difference between no filter and the multi-coated, perhaps because the lens has more contrast to start with at this focal length. Probably not a very noticable difference in all but the strongest contra-light situations (like this contrived one), but nonetheless there is a difference.
Well the Tiffen isn't doing so well still. Obviously a big loss of contrast. RGB is 78,77,68.
Just for giggles the light box was turned off and the identical exposure taken (no filter if it matters) and a RGB value of 0,0,0 is obtained.
Now it was time to test some flare. A halogen desk lamp was set just outside the field of view (you'll see it better in later shots) and once again a series of exposures were taken. These were all done at 200/4.0. It is a bit harder to get a read on the central RGB value since the flare varies over the dark box. These measures should probably be taken as plus or minus a count. Here, with no filter, we get 156,82,46. This is just the reference value for flare for this particular situation with this lens.
Visibly the Hoya S-HMC does suffer a bit more flare in this drastic situation. The RGB value here is 165,90,50. Certainly noticeable but by no means objectionable, especially considering the severity of the test.
My, does the Tiffen suck up a storm here. RGB value is really hard to pick because of the severe flare across the dark box, but picking the center the same as on the other two shots the RGB is 237,166,101. Ouch.
Here from the same shooting position as the previous shots the camera is zoomed out to 70mm/4.0 to get the light source in the field of view. This allows us to examine the individual internal reflections. There will be no specific measurements for this set (it'd be hard to decide what to measure) but there is much to be learned from visual inspection. This is the reference image with no filter.
Here we have the Hoya S-HMC and we can see a slight increase in the tonal levels of the major reflections. Nothing objectionable (pretty amazing in my opinion) but certainly a bit more than without any filter at all.
Oh my, the poor Tiffen. Not only did everything get a lot brighter but notice the whole new reflection in the upper right of the image (a rather large yellow disk).
Now for the same total exposure we stop the lens down to F/16. This allows us to sort out the individual reflections much more easily. These are fun to look at full size (you can click on them to get to a larger size or even the originals). Circular reflections occured between elements in front of the aperature stop. Octagonal reflections occur past the aperature stop (that's the shape of the aperature stop you're seeing). What exactly is causing the greenish flower petal reflections near the center is beyond my knowledge of optics, but it is pretty cool whatever it is. Anyway, this is the reference image without any filter.
Here we have the Hoya S-HMC. Again, a tiny bit more flare, noticable side by side or switching between the two images. Nothing awful, again I'm pretty impressed all things considered.
Well, I guess the Tiffen was cheap at least. This is pretty cool though, notice that not only is the flare way brighter but a whole new family of internal reflections has show up. If you've got one of these crappy UV filters I suggest you remove it before pointing it at anything bright!
So somehow in various forums the notion that coatings really aren't that important on polarizers has come up. I couldn't think of any reason why this would be true and as luck would have it I had coated and uncoated (as far as I can tell from Tiffen's marketing material) 67mm polarizers. Now, in the interests of full disclosure I should mention that one is a Hoya S-HMC circular polarizer and the other is a Tiffen linear polarizer (I used it on 4x5 cameras that don't autofocus or expose). I apologize in advance that this is not a true apples to apples comparison, but one should note that the circular polarizer has an additional layer in it to convert the linearly polarized light to circularly polarized light so if anything it should do worse. NOTE: Please see the addendum at the end of this article. It turns out that differing levels of attenuation between the two types of polarizers slightly reduce the measured contrast loss for the uncoated filter by perhaps 10% from what the numbers here suggest. There is still a lot of contrast loss all the same. I've not done any "reference" no filter shots here because of the exposure difference of a polarizer. I have ensured that the polarizers were oriented the same though. So here's the Hoya S-HMC at 70/4.0 and the RGB is 83,83,85.
And here is the Tiffen. Hmmm, looks pretty darn awful, I'd say optical physics didn't take a break for lunch just because these are polarizers. The RGB is 143,141,131. Awful. I guess my Hoya S-HMC was eighty dollars well spent (or at least better spent?).
Here's the flare test at 70/4.0 with the Hoya S-HMC circular polarizer. RGB in the center is 180,135,93.
And the Tiffen is looking its usual awful self. Again, a bit hard with the RGB value due to the changing flare across the dark box, but again we pick the center and get 241,191,134. Not a good show at all.
To conclude the polarizer test we again get the source into the image. This time the shot was only at F/11 (exposure was already up to 30 seconds). This is the Hoya S-HMC.
And here we have the Tiffen. I'm pretty sure my light box is in the image somewhere...
The Conclusion
Here are the results of the 200mm contrast test side by side. The 200mm test was choosen since it appeared to have the largest difference between the no filter and multicoated filter cases. I'm interpreting this as the 70-200 having a bit more contrast to start with out at 200, but I could be wrong. It is worth noting that in the 70mm case it is impossible to distinguish the no filter and Hoya S-HMC filter squares. It appears that in very strong contra-light and flare situations that a good multi-coated UV filter holds up pretty darn well. It does in fact cause a bit of loss of contrast, but nothing awful. In a more normal day to day shot without a strong contra light source there'd be no noticable difference. An uncoated filter on the other hand is an abomination if bright light gets near it. Similarly for polarizers coatings really are a good idea. Sorry to say I think we'll just have to pay for performance. What does this mean for me? It means my Hoya S-HMC UV filters are staying on my glass. For all but the most extreme cases they have no apparent impact on flare or contrast. I'll probably take them off in the rare cases I'm dealing with a serious contrast issue but most of the time they'll stay on keeping the dust out and, in not too long for me, salt laden marine iguana sneezes off the front optic. In addition I'll be trying no filter/S-HMC filter comparisons of real world shots as I come upon them. Others have reported some AF issues with UV filters. I've yet to experience but will be on the look out. If you'd like to read Hoya's marketing-speak about their coatings you can go here. Remember, in this test both Hoya filters were Super HMC filters. I had some issue verifying that the Tiffen filters were in fact uncoated (as opposed to single coated). I'm leaning towards uncoated since no where could I find Tiffen making claims on any coating at all and I'm sure they'd take the opportunity to point them out if they had them. In addition a FAQ on the B&H website specifically refers to the Tiffen filters as uncoated. What about other brands? Well I'd expect high quality multicoated filters in general to perform similarly. Of course there is more to a filter than just its coatings (e.g. construction and color shifts) and many people have religous affinities for various brands. I'm not plugging Hoya here, I just happened to like their price point for a good multi-coated filter. As far as the Hoya HMC filters (that is not "super") go I don't have any to test, but based on the Hoya marketing babble I'd expect them to be superior to the uncoated by a wide margin and perhaps a little worse than the Super's just based on their slightly higher reflection numbers. Hopefully you've found this helpful.
Addendum
White RGB Values from UV Test
Curiosity was expressed by some as to what the white RGB levels of the UV filter tests were in order to see if there was any difference in attenuation in the three cases. The assumption had been that any attenuation would be minimal and that any that existed would serve to emphasize the progression of loss of contrast since obviously the no filter case will not attenuate at all, the multi-coated filter should attenuate minimally due to its coatings and the uncoated should attenuate about 8% at most due to reflection. Well, it certainly never hurts to check! The original images were of course exposed as a contra light situation in which the background is overexposed so there is no useful data to be found there. Instead I made three more exposures at a lower exposure in order to get the white level within the dynamic range of the camera. In this image there are white squares from each of the three cases. The results are as expected, there is very slight attenuation by the multi-coated filter (Hoya S-HMC) and measurable though not significant attenuation by the uncoated (Tiffen). It is also worth noting the slight color shift of the Tiffen as a result of a slightly higher attenuation in the blue channel. This should not be held against the Tiffen as it is a Haze(2A) filter as opposed to a UV(0) filter and rolls off a little bit more into the blue part of the spectrum (i.e. at a longer wavelength). The average RGB values are: No filter: 240, 231, 222 Hoya S-HMC: 238, 230, 219 Tiffen: 237, 225, 209 Again, this data only reinforces the orignal conclusions drawn regarding relative contrast since contrast is really a measure of white level over black level. The orginal assumption was that the white levels would be nearly identical, they are in fact very close but are different in a manner that emphasizes the loss of contrast for both filters.
White RGB Values from Polarizers
While we expected the UV filters to have very little difference between them on the white side, upon further consideration we really should check the polarizers. Polarizers are known to have different color shifts, and especially as the circular polarizer may have additional attenuation as a result of its conversion from linearly polarized light to circularly polarized light it would definately be worthwhile to check. Here are white squares for each filter, the average RGB values are: Hoya S-HMC (circular, multi-coated): 230, 220, 210 Tiffen (linear, uncoated): 236, 234, 222 Well, our suspicions are realized. There is a color shift and as we expected there is slightly more attenuation from the circular polarizer. It is important to note that whereas in the UV tests the differences in white levels actually reinforced the results from just measuring black levels, in this case the opposite is true. For the polarizers these white level differences slightly reduce the contrast loss between the two filters. It is worth noting, however, that after accounting for this small difference in white level there is still a significant contrast loss (the white levels differ by about 6 while the black levels by about 60). The flare results still speak for themselves. But again, checking is always a good thing!