The good folks at Caprock Development were kind enough to loan me a full set of their Caprock Anti-Moire filters (1.0, 1.4, 2.0, and 2.8) for testing on a 7D, to find out what works and what doesn't, and why we hear conflicting reports about these filters.
I spent quite a bit of time with these filters, and I think I came to some pretty solid conclusions that can hopefully clear up the confusion.
First, let's get to what and why: the Caprock filters are designed to combat moire and aliasing. They're labeled "re-screener filters" and they come in various configurations; 3x3 squares or 67mm screw-in. The set I tested was the 67mm versions.
I don't think it's appropriate to refer to them as different "strengths". I've heard people say that a 1.4 is "stronger" (i.e., more blur) than a 2.0, but in my testing, that's simply not the case. The Caprock filters have a wavelet pattern on them that makes them look like... well, like polka dots. Or (yes) a cheese grater. The filters let certain frequencies pass through unmodified, and they blur or mitigate other frequencies. The 2.8 has the pattern of the largest circles, the 1.0 has the tightest-circle pattern.
What I found is that you have to match the appropriate size circles to the appropriate focal length of the lens you're using. Some of the filters worked miracles on removing aliasing on one lens, and were absolutely blurry picture-ruining disasters on other focal lengths. You cannot just buy one Caprock anti-moire filter and expect it to work for all lenses, it does not and will not work that way. And I think it would be a profoundly bad idea to try to use one of these filters on a zoom lens. You have to match the size of the circles with the focal length of the lens -- get it right, and it works very very well. Get it wrong and it's horrible. A zoom lens provides so many opportunities to get it wrong, that I think you're better off not using the anti-moire filters at all, rather than trying to use them on a zoom.
So, what works and how well does it work? I tested using a set of Zeiss ZF lenses. These are some of the sharpest, crispest SLR lenses available; Zeiss actually offers these exact same lenses as cinema lenses as well (the Zeiss Compact Primes). Wide open they can be soft, but stop them down to f/5.6 or f/8 and they're breathtakingly razor sharp.
My methodology in testing was to use the absolute sharpest image I could get, because aliasing happens when the lens is passing too much detail on to the chips. Aliasing (and moire) is when false detail gets through to the chips and contaminates the image. The goal would be to pass as much "real" detail, while blocking as much "false" detail, as possible. That would give you ultimately-sharp, properly-resolved, contamination-free footage.
I used the resolution chart in my DSC Labs Cambook 6 for the image to test. I have other, better charts, but those didn't seem right for this. With the Cambook chart, defined patterns of lines are shown without a wedge. I wasn't trying to distinguish the ultimate point of res failure, I wanted to see what would allow the most clearly, cleanly sharply resolved block of lines at the highest res, while preventing aliasing (which is extremely easy to observe in a block of lines). The Cambook 6 chart was perfect for this.
An extraction of what the chart looks like with the bare lens, and how the aliasing manifests itself, is visible in this small video file:
As a base for measurement, I shot the chart with the bare lens at f/5.6, and very mildly wiggled the camera around. Those blocks on the chart that stayed cleanly resolved with no bizarre line patterns or color contamination were considered "cleanly resolved"; if a block started showing weird patterns or color contamination, they were "aliased". With the bare lens the camera clearly and cleanly resolved the 500H and 400V blocks; there was always aliasing in the 600H and 700H blocks, mild aliasing in the 500V and extreme aliasing in the 600V, and 700V blocks.
I used a set of six lenses for testing: ZF 18mm, 25mm, 35mm, 50mm, 85mm, and 100mm. Each lens was set to f/5.6 and each was hyper-crisply focused on the chart. I attempted to frame the chart equivalently on each shot, but that wasn't possible with the 18mm; I literally couldn't get the full chart framed up and focused that close on such a wide angle, so in that case I just did the best I could. I also put the chart on a slight angle, just to be sure that the footage wasn't being unfairly influenced by lines just perfectly lining up with sensor pixels.
Here are the results:
Caprock 1.0 - fully resolved 500H, 400V but still had lots of moire present. Not an ideal solution, but the best of the Caprocks for this focal length. Very flat contrast, you'll have to boost the contrast and sharpening in-camera (or in post) to get pleasing images.
Caprock 1.4 - lots and lots of moire. Kind of pointless to use the filter, as there's still so much moire. It's better than the naked lens but not by much.
Caprock 2.0 and 2.8 -- horrible moire. Basically pointless to use these filters with this lens.
Caprock 1.0 -- worked very well. Resolved 500/400 cleanly. Very mild aliasing.
Caprock 1.4 -- worked okay, but lots of aliasing beyond 500 lines.
Caprock 2.0 -- bad aliasing
Caprock 2.8 -- almost ineffective, looked like the bare lens
Caprock 1.0 -- worked great. Flat contrast but plenty of real detail with very little aliasing
Caprock 1.4 -- very aliased
Caprock 2.0 and 2.8 -- almost ineffective
Caprock 1.0 - NO vertical aliasing, but also blurs the image. There's a pretty significant loss of true detail; it resolves around 400h/300v. There's a little aliasing on 600 lines or more in the horizontal but none on the vertical. I wouldn't use this filter/lens combination unless you're okay with standard-def images.
Caprock 1.4 - best for this focal length. Cleanly resolves 500h/400v, but it does have some significant aliasing on the vertical. It's very clean on the horizontal, but the vertical aliasing is bad on 600+.
Caprock 2.0 - bad aliasing
Caprock 2.8 - Performed very much like the 1.4! This is interesting, two different circle sizes performed very comparably. There's a little bit more horz aliasing with the 2.8, so I think the 1.4 is the better filter for a 50mm, but it's very close.
Caprock 1.0 - horrible. Very very blurry, it eliminates all aliasing but it also only resolves 200h/200v. Useless.
Caprock 1.4 - very clean with zero aliasing, but it loses some detail. It resolves 400h/400v.
Caprock 2.0 - cleanly resolved 500/400, but noticeable color aliasing on 500V or higher
Caprock 2.8 - best for this focal length. Resolved 500/500 with no color aliasing.
Caprock 1.0 and 1.4 -- useless on this focal length. Extraordinarily blurry.
Caprock 2.0 - pretty much ineffective at this focal length; looks just as aliased as the bare lens
Caprock 2.8 - resolves 400/400 with no aliasing. Clearly the best filter size for this lens, but not really a good solution because it does sacrifice some horizontal resolution.
So, there you have it -- the Caprock filters are good products, but only if used properly, with the right filter pattern being used on the right focal length lens. I am certian these results will apply to any APS-C-sized sensor camera; I would GUESS that the same would apply to the full-frame or smaller-frame cameras but am not certain on that.
If you want to see what a difference the right filter at the right strength can make, check out this video:
That's the 85mm lens with the 2.8 filter installed. Compare that against the bare lens and you'll see that it pretty much retains all the "real" detail, but gets rid of the contamination in the aliased sections. With this particular video I did boost up the sharpness in post some; the Caprock filters all cause a flattening of the contrast in the image and it makes it look "softer" than it really is; a little sharpening is necessary to get a pleasing image.
So, regarding the individual filters and what they worked best on:
Best on: 25mm, 35mm
Okay on: 18mm
Bad on: 50mm, 85mm, 100mm
Best on: 50mm
Okay on: 85mm
Bad on: 18mm, 35mm, 100mm
Best on: none
Okay on: 85mm
Bad on: 18mm, 25mm, 35mm, 50mm, 100mm
Best on: 85mm
Okay on: 50mm, 100mm
Bad on: 18mm, 25mm, 35mm
Results 1 to 10 of 25
02-23-2010 11:38 AM
- Join Date
- Sep 2003
Last edited by Barry_Green; 02-23-2010 at 12:21 PM.
02-23-2010 01:19 PM
Cool stuff. Thanks very much for this breakdown.
02-23-2010 04:14 PM
Thank you for doing these tests, Barry. Very helpful information! (Maybe your post should be a sticky or article?)
02-23-2010 07:44 PM
Last edited by ROCKMORE; 02-23-2010 at 07:53 PM.
02-23-2010 08:38 PM
- Join Date
- Sep 2003
I would assume that the focal length comparison would hold across brands.
But I would not be confident at all that one type of filter would hold for the whole zoom range. Look at the results for the 50mm; the 1.4 was great, 2.0 was awful, and 2.8 was very good. That's not the kind of performance that makes me think one particular type of filter would hold across an entire zoom range. They're not different "strengths", it's not like a normal blur filter.
But, hey, with that said, maybe it would, someone would have to test it to see for sure.
02-23-2010 10:14 PM
If one of those lenses is what you already use as a daily driver, it may save you is some situations.
I'll be interested in seeing if the 550D has the same level of moire and aliasing trouble as the existing 7D & 5D models. Should be a lot of 550D tests out by the weekend. B&H must look like a ticket line for a Stones tour. Thanks for taking the time to test these filters. The list is a good starting point to get in the ball park before ordering.
Last edited by ROCKMORE; 02-23-2010 at 10:26 PM.
03-24-2010 09:09 PM
- Join Date
- Jun 2005
Thanks for the test to start me in the right direction.
What resolution were you shooting at?
Since the 720 mode shoots at up to 60 fps, my assumption would be that it has to skip more lines than 1080 in order to read out that much data.
I could be wrong, but if this were true then that would mean that the 720 mode would need a different value Caprock filter to cover the wider gap of skipped lines than the 1080 mode would at the same lens length.
And if so, then I wonder how the 640 mode functions as well.
What are your thoughts on this? What resolution did you shoot at for your tests?
Last edited by FatBird19; 03-24-2010 at 09:18 PM.
03-25-2010 08:26 AM
- Join Date
- Feb 2009
great post... thanks