Do some lenses cause more aliasing than others?

JustinBrown

Well-known member
I recently rented a canon 28mm 1.8, and shot some b/w low light tests. On previous tests with my 50mm 1.8 I saw little to no serious aliasing, but was blown away by the jagginess of the footage filmed with the 28mm. All camera settings other than the monochrome were according to hoyle. Sharpness/contrast were all the way down. Can anyone help me figure out what the heck happened? Or is this normal?

Screenshots:
http://www.flickr.com/photos/94006943@N05/8576718949/in/photostream/
http://www.flickr.com/photos/94006943@N05/8577819978/in/photostream/

Thanks for you help!
 
I'm not an optical physicist or what-ever those guys are, however, I think aliasing is caused by pattern complexity in relation to the camera's magnification or reproduction of that design. A very wide patterned object may not alias up close, but maybe at 30feet away the design becomes small enough to start tricking the sensor pixels.

The same goes with lenses. A complex fabric design may not alias on a 100mm lens, but when you widen out at a 50mm, the complex pattern may 'shrink' into a tighter pattern and thus become small enough in the camera to start doing strange things.

I'm willing to bet, if you increased your distance by double to the object and used that 50.1.8mm, you'd get more aliasing. I'd also say if you walked forward with your 28mm by reducing your distance to the object by around half, you'd see less. Of course a razor sharp lens will likely be able to pick up a complex pattern in the distance at greater detail thus alaising might be more with super sharp lenses at times, it's an interesting thought.

My .02$


So I don't think it depends on the brand of lenses or something like that.... I think it has to do with the relationship of the complex pattern/subject being shot and it's magnification/size on the imager.
 
I'm not an optical physicist or what-ever those guys are, however, I think aliasing is caused by pattern complexity in relation to the camera's magnification or reproduction of that design. A very wide patterned object may not alias up close, but maybe at 30feet away the design becomes small enough to start tricking the sensor pixels.

The same goes with lenses. A complex fabric design may not alias on a 100mm lens, but when you widen out at a 50mm, the complex pattern may 'shrink' into a tighter pattern and thus become small enough in the camera to start doing strange things.

I'm willing to bet, if you increased your distance by double to the object and used that 50.1.8mm, you'd get more aliasing. I'd also say if you walked forward with your 28mm by reducing your distance to the object by around half, you'd see less. Of course a razor sharp lens will likely be able to pick up a complex pattern in the distance at greater detail thus alaising might be more with super sharp lenses at times, it's an interesting thought.

My .02$


So I don't think it depends on the brand of lenses or something like that.... I think it has to do with the relationship of the complex pattern/subject being shot and it's magnification/size on the imager.

Thanks Ryan, that makes a lot of sense. We shot some medium distance footage and the aliasing was well within the realm of expectation. Is there anything that can be done that would allow wider angles while minimizing the aliasing? More lighting? Filters? A shallower depth of field?
 
The relationship to lenses and aliasing is simply this: IF a camera is prone to aliasing, then using a sharper lens will cause aliasing to be more evident.

Aliasing happens when the sensor cannot handle as much detail as is being thrown at it. The sharper the lens, the more detail it will transmit to the sensor.

Now, this might lead some to thinking that you should use softer lenses, to avoid the problem -- and, in fact, one of the recommended ways to minimize aliasing is to slightly defocus the image. Which, while it will work, is -- crazy, isn't it? If you've got a fundamentally defective image processing chain (the aliasing sensor), does it really make sense that to try to cope with it, you have to cripple the lens performance? But, really, it's the only way, until you can take the step that will truly solve it properly -- move up to a proper-performing sensor. And that's not likely going to be reality in the DSLR price bracket for several years at least; there's only so much that can be done at the price point that exists with today's technology.

So -- assuming that you're stuck with a line-skipping sensor, understand that that type of sensor is going to be very sensitive to patterns of repeated detail (which will show up as ugly aliasing), and the only "solutions" are to change the frequency of the pattern of the image so that it doesn't conflict with the sensor's pixel-skipping pattern (by moving closer or further away), or, to really prevent the problem in the first place, you'd have to soften the detail down so that the sensor can cope with it. And that means -- yes, using worse-quality lenses, or de-focusing the image slightly, until the aliasing goes away.
 
The relationship to lenses and aliasing is simply this: IF a camera is prone to aliasing, then using a sharper lens will cause aliasing to be more evident.

Aliasing happens when the sensor cannot handle as much detail as is being thrown at it. The sharper the lens, the more detail it will transmit to the sensor.

Now, this might lead some to thinking that you should use softer lenses, to avoid the problem -- and, in fact, one of the recommended ways to minimize aliasing is to slightly defocus the image. Which, while it will work, is -- crazy, isn't it? If you've got a fundamentally defective image processing chain (the aliasing sensor), does it really make sense that to try to cope with it, you have to cripple the lens performance? But, really, it's the only way, until you can take the step that will truly solve it properly -- move up to a proper-performing sensor. And that's not likely going to be reality in the DSLR price bracket for several years at least; there's only so much that can be done at the price point that exists with today's technology.

So -- assuming that you're stuck with a line-skipping sensor, understand that that type of sensor is going to be very sensitive to patterns of repeated detail (which will show up as ugly aliasing), and the only "solutions" are to change the frequency of the pattern of the image so that it doesn't conflict with the sensor's pixel-skipping pattern (by moving closer or further away), or, to really prevent the problem in the first place, you'd have to soften the detail down so that the sensor can cope with it. And that means -- yes, using worse-quality lenses, or de-focusing the image slightly, until the aliasing goes away.

Barry - What initially caused my concern wasn't detailed patterns, but surface edges. Please see my screenshots for a better idea of what I am talking about. I think Ryan covered what was concerning me, but I would still appreciate any idea's or feedback on how to tackle that problem.

Thanks!
 
It's the same thing; aliasing manifests itself in many ways, whether pixelized lines or edges or moire patterns. But it all comes down to the same root cause -- too much detail being fed to an undersampling sensor. And the solutions are still the same -- either filter out some of the detail, or use a sensor that can handle the detail being fed to it. Since not all of us (okay, quite few of us) have the luxury of upgrading the camera head to a higher-quality sensor, that still leaves you with the same challenge of trying to drop the detail down to where the sensor can handle what it's seeing.

In the case of the archway, you have to understand what the root cause is -- the sensor in DSLRs is a bodge, a hack, a kludge, a "cheat", and it simply just does not work very well for handling real detail. So you either have to lower the detail (by throwing that archway out of focus), or use a camera that has a real sensor that's designed for the job you want it to do. I guarantee you you'd be infinitely happier with your edges on an F5, F55, Red Epic, or Alexa. Those cameras are using real sensors that are designed for the job of shooting motion images, and have adequate resolution and proper OLPF filtering to render the images properly.

A DSLR is designed to shoot 18-megapixel stills. The process to get it to shoot 2-megapixel video is sketchy and results in exactly the type of problems you're asking about. It is, in short, absolutely inherent to the concept and notion of shooting video on a DSLR using today's technology. Maybe three or five years from now the technology will have advanced to the point where they can put proper scaling on a DSLR sensor and get proper image reproduction, but as of today's technology, it simply doesn't exist.

So, again, you're left with the only possible fixes -- either upgrade to a sensor that can handle the job, or artificially drop the detail level down to a point where the DSLR sensor won't choke on it. And that means de-focusing, through either actually forcing the image out of focus, or narrowing the depth of field such that the offending region is now properly blurred, or shooting at a diffraction-inducing aperture such as f/22 so that everything is forced slightly out of focus below the aliasing threshold.
 
Or is this normal?
Going back to the original question -- yes, it's absolutely normal. If nothing else, it's the deeper depth of field of the wider-angle lens which is causing you to notice the problem. The wider the lens, the deeper the depth of field, and the more that is in focus. You'll see less aliasing from your 50mm lens because the background is always softer in focus. The sharper background of the wider lens causes you to notice the aliasing problem more.
 
A DSLR is designed to shoot 18-megapixel stills. The process to get it to shoot 2-megapixel video is sketchy and results in exactly the type of problems you're asking about. It is, in short, absolutely inherent to the concept and notion of shooting video on a DSLR using today's technology. Maybe three or five years from now the technology will have advanced to the point where they can put proper scaling on a DSLR sensor and get proper image reproduction, but as of today's technology, it simply doesn't exist.

I think it could be mitigated with better DSP processing that 'subsamples' the sensor, filtering for jaggies, to produce the 1920x1080 2-d image. The problem with a 'fixed' optical Low Pass Filter... is it would screw up the stills at their max resolution... No doubt the OLPF is designed to cut of higher spatial frequencies for that resolution.

Of course, DSLRs have significant limitations on their processors due to power/heat issues...

And the question is... if a camera doesn't have the need for all the DSLR still 'works', why is the price for such a device, 2x or 3x the DSLR price, such as the AF100 or the C300.

I've not kept up on the Black Magic Camera, as it's price is also out of my 'price range' of free or less... but it seems to me that camera should be lower in price than it is...
 
The price is higher because the sensor and processing is designed for the job at hand. DSLRs don't do everything dedicated cinema cameras do, not even close, not by a hundred miles. So why would leaving off DSLR features, make a cinema camera cost less?

If the DSLR did everything the cinema camera did, and also took stills, then you might have a point. But when the DSLR does a lousy job of pixel-skipping and scaling, then how can you even begin to compare that against a designed-for-the-purpose dedicated cinema camera like a C300?
 
...and consider that for every C300 sold, there are probably 100,007 DSLRs sold (it may be 100,008 though). Market and scale are big factors.
 
The low cost solution to getting a better performing sensor in the aliasing/moire arena for video is the Nikon D5200. The second video is just to show the picture styles...but also read the description below. Visit the Nikon forums here...this camera is flying under the radar, but most have been very impressed. Toshiba has created a very nice sensor.

http://vimeo.com/58959901

http://vimeo.com/59767574
 
Going back to the original question -- yes, it's absolutely normal. If nothing else, it's the deeper depth of field of the wider-angle lens which is causing you to notice the problem. The wider the lens, the deeper the depth of field, and the more that is in focus. You'll see less aliasing from your 50mm lens because the background is always softer in focus. The sharper background of the wider lens causes you to notice the aliasing problem more.

Thanks for clearing that up(for what must be the umpteenth time). I'd read some of your comprehensive posts on this subject earlier when I was banging my head against the wall trying to determine the cause of the jaggedness. The real problem here, for me, is lack of experience. This was the first real test drive I've done with my camera outside of filming landscapes with the kit lens and I feel I may have over-estimated the resolving power of the sensor based on those results(all were brightly lit, with a setting of f/14+). I also had not realized the greater depth of field of wide angle lenses, and only went off of the experience of the 50 1.8 which worked great, with only minimal aliasing. Since we were shooting primarily in the 2.2 to 2.8 range I just assumed I could transpose the results of the 50mm to the 28mm. Phew! This all helps a great deal.

To summarize:

1. Soften the focus by shortening the depth of field.

2. Shrink the aperture to increase the diffraction.
 
Thanks for clearing that up(for what must be the umpteenth time). I'd read some of your comprehensive posts on this subject earlier when I was banging my head against the wall trying to determine the cause of the jaggedness. The real problem here, for me, is lack of experience. This was the first real test drive I've done with my camera outside of filming landscapes with the kit lens and I feel I may have over-estimated the resolving power of the sensor based on those results(all were brightly lit, with a setting of f/14+). I also had not realized the greater depth of field of wide angle lenses, and only went off of the experience of the 50 1.8 which worked great, with only minimal aliasing. Since we were shooting primarily in the 2.2 to 2.8 range I just assumed I could transpose the results of the 50mm to the 28mm. Phew! This all helps a great deal.

To summarize:

1. Soften the focus by shortening the depth of field.

2. Shrink the aperture to increase the diffraction.

Well you can't do both!




Most of my moire/aliasing issues are found in complex patterns on subjects clothing, and thus my explanation earlier focused on simply staying away from those patterns and how even acceptable patterns upclose can become an issue when further away or using different focal length 'magnifications.'

Barry has a good point with focusing. Small defocusing can help a ton, if you can accept a slightly soft image. Maybe even using filtration such as Classic Soft, ProMist, BlackProMists, etc. can help a ton too.

The suggestion of using less sharp lenses is also a great point, but in my opinion it only theoretically works, not so often in practice. I say this because almost all modern lenses of professional caliber are plenty sharp enough! Unless using some very very vintage glass, big vintage zoom, or something in that range, most lenses are plenty sharp to aggravate aliasing. In fact, unless very vintage or outright damaged, I don't see myself finding many lenses soft enough to hide aliasing, even a little. Perhaps some really fast and old lenses wide open?

Maybe it's been awhile since I've used soft lenses, but I find kit lenses with NEX cameras, which are like $200 to be plenty sharp! And they are cheap glass!

I think shooting shallow and watching for complex patterns is a good way to start. Maybe using a classic soft filter too. Most of my aliasing battles are on higher end cameras, so I'm not used to seeing it everywhere.
 
Last edited:
As Ryan has pointed out, pretty much all lenses, even cheap lenses have plenty high MTF graphs at low spatial frequencies.
For a full frame DSLR camera shooting fullhd you are interested in spatial frequencies up to around 13-14 lp/mm, anything above this is only food for aliasing. (With lineskipping and debayering the relationship is more complex, but lets not get into details.) The point is, to get a low MTF result (and low MTF = lower local contrast = less apparent aliasing on edges) you need a real soft lens, which is a special fx lens. Soft lenses may have MTF results as low as 50% even at low frequencies, and the graph stays more or less flat with frequencies going up. This is really just a technial description for flare/veiling glare and highlights locally bleeding into shadows, effectively lowering edge contrast. Note that edge definition is not reduced, only edge contrast. One other way to get this is to use a Softar (or a similar type) filter. This may or may not be practical depending on your target look. Shooting in a low contrast picture profile may also help a little bit, but affects global contrast.

You can also lower local contrast (i.e. clarity) and/or apply some edge blurring filter in post.
 
its such a hit and miss thing. pretty much all canon DSLRS below the 5d3 are pretty bad about it. the GH2 had nearly no problem with it. but the GH3 does.
from what ive heard (and seen to an extent) all the nikons did as well. butt the d5200 looks just a bit more prone than the GH2. may have to have a look into that one...
 
Well you can't do both!




Most of my moire/aliasing issues are found in complex patterns on subjects clothing, and thus my explanation earlier focused on simply staying away from those patterns and how even acceptable patterns upclose can become an issue when further away or using different focal length 'magnifications.'

Barry has a good point with focusing. Small defocusing can help a ton, if you can accept a slightly soft image. Maybe even using filtration such as Classic Soft, ProMist, BlackProMists, etc. can help a ton too.

The suggestion of using less sharp lenses is also a great point, but in my opinion it only theoretically works, not so often in practice. I say this because almost all modern lenses of professional caliber are plenty sharp enough! Unless using some very very vintage glass, big vintage zoom, or something in that range, most lenses are plenty sharp to aggravate aliasing. In fact, unless very vintage or outright damaged, I don't see myself finding many lenses soft enough to hide aliasing, even a little. Perhaps some really fast and old lenses wide open?

Maybe it's been awhile since I've used soft lenses, but I find kit lenses with NEX cameras, which are like $200 to be plenty sharp! And they are cheap glass!

I think shooting shallow and watching for complex patterns is a good way to start. Maybe using a classic soft filter too. Most of my aliasing battles are on higher end cameras, so I'm not used to seeing it everywhere.

Of course! I should've added an "or" in there. It seems like a case of "pick your poison", and counter intuitive to invest in a bunch of lesser quality lenses to dumb down my image. Another thing I keep telling myself is "if I don't point it out, no one will probably notice". Heck, I didn't even notice the small amounts I had recorded until I got paranoid staring at the very obvious escalators running through some of the images in the later 28mm footage. Now it's all I can see, which on one hand is cool, awareness is half the battle, but on the other...soooo annoying.

Thanks again...so much to mull over now...
 
Today, that statement that "all DSLRs have aliasing/moire" is no more true than "all video cameras have aliasing/moire".
Specifically, the D5200 doesn't seem to have more aliasing/moire than, say, an EX1: if you're adamant to make it appear, it will show up, but it's going to be a feat.
 
wide lenses will alias more than long focal lengths because out of focus parts of the image will hide moire and aliasing
 
every single digital camera (still or video) out there, alexa, red, f5 can moire and show aliasing
 
every single digital camera (still or video) out there, alexa, red, f5 can moire and show aliasing
Thank you for bringing that up. I've seen quite a bit of both (maybe not to the same extent as some of the mentioned DSLRs) from those expensive pro cameras (from the AF100 - F3). I think the first time I noticed aliasing was from some EX-1 footage.
 
Back
Top