Can a UV filter affect IQ (sharpness) on a lens?

[RC]

I have a 17-55 IS and use it with a B+W MRC UV Filter which is always on. I still tend to get dust particles inside, this lens is notorious for this. I had it sent to Canon one time for cleaning (and to replace the IS motor, total was like $130). There are some small dust particles again inside but it never shows up in the pictures.

Interesting. I sure would like to know what the differences are for those who get lots of dust and those who get none.

- environment, wind, dust, humidity, geography
- filters, lens caps off
- amount of zooming
- amount of use
- manufacturing, serial numbers

Obviously too many variables to really know

 

[Loswr]

A UV filter always affects image quality, including sharpness. It lowers performance in sharpness, contrast, transmission, etc., by at least 5%. For some filters, more than 10%. That's doesn't sound bad considering that the Sony SLT cameras absorb 33% of light (transmission only), but here we are talking about a 5% or more reduction in every optical characteristic of the lens, not just transmission.

It gets worse the larger the area of the filter. So for 18-55mm lenses it's not really important. But for a 300mm f/2.8 lens, it would be ridiculous.

I will respectfully disagree. I've shot with/without tests using a B+W MRC UV filters and an ISO 12233-based test chart which costs more than some L-series lenses, and detected no difference in sharpness. Granted, that was under ideal lighting conditions. I have also tested the effect on flare, and there is an increase. Although my flare testing was a 'worst case scenario' - very bright light in the extreme corner of the image, in an otherwise dark room (literally, a darkroom) - the increased flare is sufficient to reduce contrast. Reduced contrast will manifest as a reduction in sharpness, so in that sense, with side lighting or backlighting, a filter may cause a reduction in sharpness. I'd think it will fall short of a 5% loss in most cases.

Of course, as you know the 300/2.8 does not take a front filter, only a drop-in 52mm filter. Interestingly, while you can choose to leave off a front filter, the supertele lenses actually require a filter. Canon states you should keep a clear glass filter in that drop-in holder (check the manual) because the optical design of the lens incorporates a filter in the holder.

 

[Mt Spokane Photography]

Filters always affect IQ, even air affects IQ, its just a matter of how much. Although I have plenty of B&W and Heliopan filters, I avoid using them unless its dusty, or likely to get damage, water, salt spray, etc on the lens.

I always kept a filter on my 17-55 and after 3 years had no internal dust. That was the one exception to my occasional use of filters.

I have very few cases where the difference in sharpness would be noticible except by obsessing ofer a 100% view of the image. For a wide angle lendscape where I need all the detail I can get, I'd remove any UV filter.

 

[helpful]

A UV filter always affects image quality, including sharpness. It lowers performance in sharpness, contrast, transmission, etc., by at least 5%. For some filters, more than 10%. That's doesn't sound bad considering that the Sony SLT cameras absorb 33% of light (transmission only), but here we are talking about a 5% or more reduction in every optical characteristic of the lens, not just transmission.

It gets worse the larger the area of the filter. So for 18-55mm lenses it's not really important. But for a 300mm f/2.8 lens, it would be ridiculous.

I will respectfully disagree. I've shot with/without tests using a B+W MRC UV filters and an ISO 12233-based test chart which costs more than some L-series lenses, and detected no difference in sharpness. Granted, that was under ideal lighting conditions. I have also tested the effect on flare, and there is an increase. Although my flare testing was a 'worst case scenario' - very bright light in the extreme corner of the image, in an otherwise dark room (literally, a darkroom) - the increased flare is sufficient to reduce contrast. Reduced contrast will manifest as a reduction in sharpness, so in that sense, with side lighting or backlighting, a filter may cause a reduction in sharpness. I'd think it will fall short of a 5% loss in most cases.

Of course, as you know the 300/2.8 does not take a front filter, only a drop-in 52mm filter. Interestingly, while you can choose to leave off a front filter, the supertele lenses actually require a filter. Canon states you should keep a clear glass filter in that drop-in holder (check the manual) because the optical design of the lens incorporates a filter in the holder.

Neuro, 5% is a very small amount, but that is an old literature value for least possible amount of image quality reduction on light rays passing through an optical system (one element or a group of elements without any intervening air to glass interfaces). I tried to look that number up again, and was unable to. However, I talked to one of my best friends today, who was responsible for designing the most expensive lens system ever created at that time for the Voyager I and Voyager II space probes (an f/0.7 lens design that used a single block of beryllium valued at over $2 million, the largest one ever manufactured). He gave me his professional opinion that with current technology (coatings, etc.) the minimum effect on image quality per air to glass interface is 3%. Filters have a bigger effect on image quality than any other element in an optical system because both sides are a flat air-to-glass interface. So the image quality would be 97% of it's original quality when passing inside of the filter, and then (97%) times (97%) when passing through the other side into the lens. That leaves the image at about 94% of it's original quality.

Looking at the resulting image, it's almost impossible to say "This image looks like 94% of perfect," because the lighting conditions in the picture, etc., can all affect the image quality (clear day vs. hazy day for example).

One time there was a salesman saying that an oven was so good that the heat would never escape. That's impossible, and so is the idea that a filter doesn't affect the image.

Search the internet for some photos of top quality filters stacked on top of each other. You can clearly see from a photo looking through the stack of filters that the rims of the filters become less and less well defined as the number of filter layers increased.

For example, after 10 filters there would be 73.7% transmission with an optically perfect set of filters (.97)^10 = 73.7%.

Depending on the angle of the light ray passing through the filter, the abberations / MTF deficiency will vary, but once again, 3% is basically a minimum level of effect.

 

[Loswr]

However, I talked to one of my best friends today...He gave me his professional opinion that with current technology (coatings, etc.) the minimum effect on image quality per air to glass interface is 3%.

For example, after 10 filters there would be 73.7% transmission with an optically perfect set of filters (.97)^10 = 73.7%.

Depending on the angle of the light ray passing through the filter, the abberations / MTF deficiency will vary, but once again, 3% is basically a minimum level of effect.

Sorry, going to disagree again. First off, I'm not sure you're distinguishing between transmission loss and IQ decrement. Transmission loss with modern lens coatings is much less than 1% per interface.

Now, take your friend's value of 3% IQ loss per interface leading to a 6% loss of IQ from a filter. Roger Cicala's (lensrentals.com) tests of large filter stacks refutes that. Moreover, optically there's no difference between a curved glass-air interface and a flat one - the magnitude of refraction and reflection is the same, only the vector direction differs. So, let's look at the 70-200mm f/2.8L IS II - arguably the best zoom lens in existence. It has 19 groups, meaning 38 glass-air interfaces. At 3% loss per interface, that means absolutely no light hitting the sensor if you mean transmission, or over 100% IQ loss if you don't mean transmission. I don't think either is even close to true.

 

[Random Orbits]

However, I talked to one of my best friends today...He gave me his professional opinion that with current technology (coatings, etc.) the minimum effect on image quality per air to glass interface is 3%.

For example, after 10 filters there would be 73.7% transmission with an optically perfect set of filters (.97)^10 = 73.7%.

Depending on the angle of the light ray passing through the filter, the abberations / MTF deficiency will vary, but once again, 3% is basically a minimum level of effect.

Sorry, going to disagree again. First off, I'm not sure you're distinguishing between transmission loss and IQ decrement. Transmission loss with modern lens coatings is much less than 1% per interface.

Now, take your friend's value of 3% IQ loss per interface leading to a 6% loss of IQ from a filter. Roger Cicala's (lensrentals.com) tests of large filter stacks refutes that. Moreover, optically there's no difference between a curved glass-air interface and a flat one - the magnitude of refraction and reflection is the same, only the vector direction differs. So, let's look at the 70-200mm f/2.8L IS II - arguably the best zoom lens in existence. It has 19 groups, meaning 38 glass-air interfaces. At 3% loss per interface, that means absolutely no light hitting the sensor if you mean transmission, or over 100% IQ loss if you don't mean transmission. I don't think either is even close to true.

The lensrental page below cites that "a modern coating that is 99.9% effective, and total reflection changes to less than 2% for the simple lens and just over 3% for the complex lens." and that "Without coatings each interface would reflect about 4% of the light that reaches it."

http://www.lensrentals.com/blog/2011/12/reflections-on-reflections-the-most-important-part-of-your-lens

 

[RLPhoto]

However, I talked to one of my best friends today...He gave me his professional opinion that with current technology (coatings, etc.) the minimum effect on image quality per air to glass interface is 3%.

For example, after 10 filters there would be 73.7% transmission with an optically perfect set of filters (.97)^10 = 73.7%.

Depending on the angle of the light ray passing through the filter, the abberations / MTF deficiency will vary, but once again, 3% is basically a minimum level of effect.

Sorry, going to disagree again. First off, I'm not sure you're distinguishing between transmission loss and IQ decrement. Transmission loss with modern lens coatings is much less than 1% per interface.

Now, take your friend's value of 3% IQ loss per interface leading to a 6% loss of IQ from a filter. Roger Cicala's (lensrentals.com) tests of large filter stacks refutes that. Moreover, optically there's no difference between a curved glass-air interface and a flat one - the magnitude of refraction and reflection is the same, only the vector direction differs. So, let's look at the 70-200mm f/2.8L IS II - arguably the best zoom lens in existence. It has 19 groups, meaning 38 glass-air interfaces. At 3% loss per interface, that means absolutely no light hitting the sensor if you mean transmission, or over 100% IQ loss if you don't mean transmission. I don't think either is even close to true.

The lensrental page below cites that "a modern coating that is 99.9% effective, and total reflection changes to less than 2% for the simple lens and just over 3% for the complex lens." and that "Without coatings each interface would reflect about 4% of the light that reaches it."

http://www.lensrentals.com/blog/2011/12/reflections-on-reflections-the-most-important-part-of-your-lens

A excellent filter will not affect IQ enough for anyone to notice. NOT anyone, even fellow pro's won't be able to say "hey this photo had a Uv filter on it!"

 

[helpful]

However, I talked to one of my best friends today...He gave me his professional opinion that with current technology (coatings, etc.) the minimum effect on image quality per air to glass interface is 3%.

For example, after 10 filters there would be 73.7% transmission with an optically perfect set of filters (.97)^10 = 73.7%.

Depending on the angle of the light ray passing through the filter, the abberations / MTF deficiency will vary, but once again, 3% is basically a minimum level of effect.

Sorry, going to disagree again. First off, I'm not sure you're distinguishing between transmission loss and IQ decrement. Transmission loss with modern lens coatings is much less than 1% per interface.

Now, take your friend's value of 3% IQ loss per interface leading to a 6% loss of IQ from a filter. Roger Cicala's (lensrentals.com) tests of large filter stacks refutes that. Moreover, optically there's no difference between a curved glass-air interface and a flat one - the magnitude of refraction and reflection is the same, only the vector direction differs. So, let's look at the 70-200mm f/2.8L IS II - arguably the best zoom lens in existence. It has 19 groups, meaning 38 glass-air interfaces. At 3% loss per interface, that means absolutely no light hitting the sensor if you mean transmission, or over 100% IQ loss if you don't mean transmission. I don't think either is even close to true.

You are confusing transmittance vs. absorbance.

Let's say I take a one-inch pane of glass. I cut it in half so that it is 1/2 inch thick. I have doubled something as a result. Perhaps the amount of light that gets through??? Then I cut it in half again, so now I have doubled something again. 2*2 = 4. Does that mean that four times as much light gets through? So I could clearly get a large amount of light through if I kept on doubling the transmission, doubling the transmission, and on and on., and I guess it would be infinitely bright light when I ran out of glass to cut in half... Wrong.

But that's the mistake you are making by saying that 100% IQ loss would occur based on the 23 elements / 19 groups in the 70-200mm II lens.

You are adding the 5% 23 times and getting more than 100%, right?

The correct way to compute it is not by subtraction but my multiplication. As I did above, if there are 10 surfaces, then there is about 73% transmittance.

That is only for flat air to glass transitions.

With binoculars they have nitrogen filled optical designs, and there are similar tricks for lenses.

And another thing is that optical groups (lenses directly combined with other lenses) count as a single element.

The answer to your question is to look at the Tstop value for the 70-200mm lens.

This Tstop value is 3.4.

Calculate the number of stops lost: log_2((3.4)^2/(2.8 )^2) = .5602 stops

This translates into a 47% loss of light vs. a hollow tube of the same diameter as the lens.

We can find the amount of light lost at each of the 19 groups as follows:

1 - .47 = (1 - x) ^ 19
.53 = (1-x)^19
.53^(1/19) = 1-x
x = 1 - .53^(1/19)
x = 3.286%.

That's almost exactly what my expert friend quoted.

You can't beat someone who works for NASA, and designed lenses for the space program at age 23 back in the 1970s.

Have a great day!

 

[bycostello]

of course, it is another piece of glass in front of the lens!

 

[symmar22]

The only filter that makes sense would be a polarizer and that's due to it's special purpose. Any additional glass in the path will affect IQ. It becomes especially apparent using longer focal lengths and lenses that are already as sharp as the Hubble Space Telecope. In addition, the flat surfaces also tend to cause more internal reflections and ghosting, even ones that are multicoated. Many of Canons lenses use meniscus front lenses that are slightly curved to reduce this effect. Adding a filter on the front just defeats the engineering.

Never use a filter unless you have a specific need for it and are willing to accept some loss of IQ.

+1

Always found strange that people invest 2000$ to get the best IQ from a lens and ruin it with a 25$ piece of cheap glass....

 

[tron]

Just because you can't stand that site, doesn't mean there's not good info/posts there. I can't stand politicians but occasionally they... okay... bad example. But I think you know what I mean

;D ;D ;D

 

[Invertalon]

I just ran my 70-200 II at 70mm through Focal with and without the Hoya HD filter I use regularly.

I ran it six times total. I used live-view focus before the first shot so AF would have NO factor in these results as it stayed the same between all the tests. Three tests with filter and three without back to back to directly compare.

My three best shot results WITHOUT a filter... 2091, 2092 and 2065.

My three best shot results WITH a filter... 2010, 2045 and 2053.

Filter OFF Average: 2082.66
Filter ON Average: 2036

Difference of around 2.24% using this testing method. Would you see that in real world shooting? Probably not. But if you want to maximize IQ than filter off.

Note, even running the same test back to back shows some difference in final results... Without moving anything at all or letting the camera AF. Most likely a slightly tolerance in the analysis software. However, the results are still consistent between them.

Another side note, there is quite a noticeable reduction in light hitting the sensor with the filter on. Afterward, I did 2 set's of comparison shots with and without a filter, same exposure, WB, etc... The shot with the filter was always a hair more underexposed. I did not expect it to be as noticeable as it was. In Lightroom, both images came out to about a 8% difference in exposure with the filter (.08 slider adjustment to match the exposures).

 

[weixing]

The only filter that makes sense would be a polarizer and that's due to it's special purpose. Any additional glass in the path will affect IQ. It becomes especially apparent using longer focal lengths and lenses that are already as sharp as the Hubble Space Telecope. In addition, the flat surfaces also tend to cause more internal reflections and ghosting, even ones that are multicoated. Many of Canons lenses use meniscus front lenses that are slightly curved to reduce this effect. Adding a filter on the front just defeats the engineering.

Never use a filter unless you have a specific need for it and are willing to accept some loss of IQ.

+1

Always found strange that people invest 2000$ to get the best IQ from a lens and ruin it with a 25$ piece of cheap glass....

+1. I also found that the case and they seem to forget that the IQ of the lens will only be as good as the quality of the worst element in the optical train and the filter is part of that when added.

Anyway, the below is a link about Canon coating technology:
http://www.canon.com/technology/s_labo/light/003/03.html

Have a nice day.

 

[Zv]

I want to know more about these flourine coatings by Canon. Neuro can you explain what that does? Does it mean we no longer need clear filters?

 

[ZoeEnPhos]

The only filter that makes sense would be a polarizer and that's due to it's special purpose. Any additional glass in the path will affect IQ. It becomes especially apparent using longer focal lengths and lenses that are already as sharp as the Hubble Space Telecope. In addition, the flat surfaces also tend to cause more internal reflections and ghosting, even ones that are multicoated. Many of Canons lenses use meniscus front lenses that are slightly curved to reduce this effect. Adding a filter on the front just defeats the engineering.

Never use a filter unless you have a specific need for it and are willing to accept some loss of IQ.

+1

Always found strange that people invest 2000$ to get the best IQ from a lens and ruin it with a 25$ piece of cheap glass....

Just wish to add my personal experience - I purchase an almost brand new Canon EF 300mm f/4L IS USM and the former owner was "generous" by adding with the lens a 77mm "un-common-cheap-brand-UV-filter".
In the beginning I made the very mistake to place this filter in the end of the lens-barrel and I noticed a problem with FLARE and lack of decent contrast and somehow less colours....I then decided to NOT use any filter of any kind to this tele-lens with decent hood and saw an immediate effect of a better overall IQ in post processing when carefully examine my .CR2 RAW files before and after removing the cheap-junk-filter.

I have not attached any filter to this lens today and I am carefully using the built-in retractable lens hood with the convenience of quickly sliding the hood out/in for use/storage.

(However I am considering to buy a Nano-based slim-filter of 82mm to my Canon TS-E 24mm f/3.5L II that I am still using without any filter so far. By the way, do anyone have a suggest what would be the best choice in order to protect the wide-angel-front-lens to this quite expensive TS-E lens?)
Wishing you a wonderful coming Week-end!
/Charl

 

[Loswr]

I want to know more about these flourine coatings by Canon. Neuro can you explain what that does? Does it mean we no longer need clear filters?

The fluorine coating makes cleaning the front and rear elements easier - it's the same stuff they use on the walls in urban centers, to make graffiti removal easier. There is no protection against scratches, etc.

 

[Zv]

I want to know more about these flourine coatings by Canon. Neuro can you explain what that does? Does it mean we no longer need clear filters?

The fluorine coating makes cleaning the front and rear elements easier - it's the same stuff they use on the walls in urban centers, to make graffiti removal easier. There is no protection against scratches, etc.

Thanks!

 

[FTb-n]

The Hoya HMC will degrade image quality. That's why I switched to the Hoya HD Clear filters. They are tough and very easy to clean. Given the cost of the filter, I wanted to know if it was worth it. I did some testing with and without it and I don't see any degredation of IQ. My 70-200 f2.8L II can be glare-prone (but not as bad as the Mark I), but the filter doesn't add any glare of it's own.

 

[tron]

The Hoya HMC will degrade image quality.

Interesting. Have you used HMC Super?

 

[Mika]

The UV filter will not only affect the sharpness of a lens, but also the whole performance, including bokeh. This is specifically true for telephoto lenses, following are shots taken with Sigma 120-400 immediately after buying it (sorry, I had to resize to 70 % of the original due to attachment size limit).

I wondered about the performance of the lens as I thought it really should be better than what I got, and contacted the importer. It turned out that the cause for image quality degradation was a comparatively bad UV filter.

I actually do have some interferograms of the filter; the filter shape error can be determined from bokeh outline (diagonal lines). Without the UV filter, the lens worked much better. I still use UV filters in front of normal and short telephoto lenses, but not in the 120-400. The only reason why I still keep them on is mainly the ease of cleaning a separable glass plate.