View Full Version : AA Filters, pixels and photosites
Busted Knuckles
03-29-2016, 12:50 PM
So about year ago, there was a pretty detailed discussion on pixel size and lens resolutions. What seemed to be missed was the effect of the AA filter.
Now we have the "R" with a cancelled AA Filter and it seems AA filters are getting weaker or excluded completely.
For the pixel peeping, landscaping (non building/fine repeating pattern landscapers) the chase on usable resolution starts again?
Would a crop, high density sensor w/o an AA Filter now provide greater usable resolutions??
This was inspired by the 80d vs. 7dII thread and I didn't want to hijack that one.
Kayaker72
03-30-2016, 10:56 AM
For the pixel peeping, landscaping (non building/fine repeating pattern landscapers) the chase on usable resolution starts again?
My personal guess is that we aren't seeing a "chase" but rather a maturing market. One of the strategies companies use in a maturing market is to start chasing smaller and smaller niches in an attempt to get as much revenue as they can out of the market. I think we are seeing a number of the manufacturers do this.
So, I would say that Canon is simply chasing market niches, not necessarily "resolution." The difference is that I expect most camera bodies to stay in the 20-30 MP range with one or two entries in the high MP range rather than all bodies headed toward the high MP range.
BTW, case in point in chasing small niches...the Nikon DF.....I am not even sure what niche that is chasing...the hipster market????:rolleyes:
Would a crop, high density sensor w/o an AA Filter now provide greater usable resolutions??
So, my quick generally recollection: aliasing/moire happens when a pattern (line, etc) is smaller than can be rendered by a sensor. So the sensor records the pattern in one pixel, but not the next, etc. Then, image processor trys to make sense of a partial pattern and chaos ensues.
At least too things come into play:
the size of an image in how it is rendered onto a sensor. In the case of APS-C, it has a smaller image circle so the same sized pattern is displayed as smaller on an APS-C sensor when compared to how it is displayed on a FF sensor.
Pixel size/density of the sensor. It otherwords, how fine of detail can a sensor resolve from the image that is projected onto it.
So, if we were to compare two cameras that have the same pixel size/density, say the 7D II and 5DS (r), more patterns will potentially result in aliasing/moire with the 7DII because of the smaller image circle with APS-C sensors.
Its early, but that makes sense to me.
So, is it possible, sure, but I think there is a reason why we are seeing FF sensors have the AA filter minimized/removed before APS-C.
conropl
03-30-2016, 05:10 PM
So, if we were to compare two cameras that have the same pixel size/density, say the 7D II and 5DS (r), more patterns will potentially result in aliasing/moire with the 7DII because of the smaller image circle with APS-C sensors.
I do not think that is true. With the same pixel density, same distance to subject, and same lens (all things the same except the image circle), then the pattern would be the same. All you are effectively doing is cropping information from a FF sensor. I maybe misinterpreting your statement, but I think your fatal flaw in your logic is that you are treating the crop sensor as a means of magnification (which a lot of people seem to do). However, keeping all things the same, the crop sensor has all the same information on a pixel to pixel bases as the FF except for the pixels outside the crop camera image circle. In other words, over that same image circle area on the FF sensor, the information recorded is exactly the same as the crop sensor; and if you had moire issues from the data interpretation on the FF sensor on that portion of the sensor covered by the crop image circle, then it is going to be the same on the crop sensor.
Pat
Kayaker72
03-30-2016, 05:32 PM
....still thinking about it, but for now, I think I will stand by my logic, but let me rephrase.
Aliasing/moire is a resolution based issue. For the same image projected onto the sensor, a 50 MP sensor has more resolving ability than a 20 MP sensor (breaking MP into linear resolution). As it has more resolving ability, it is able to accurately "see" and depict finer details and will be less prone to aliasing/moire.
So, it isn't just a pixel density issue. It is a resolution issue. And because FF sensors are larger, for the same pixel density, they have greater resolving ability.
That line of reasoning still works for me. But, of course, this is not my "area."
peety3
03-30-2016, 09:59 PM
....still thinking about it, but for now, I think I will stand by my logic, but let me rephrase.
Aliasing/moire is a resolution based issue. For the same image projected onto the sensor, a 50 MP sensor has more resolving ability than a 20 MP sensor (breaking MP into linear resolution). As it has more resolving ability, it is able to accurately "see" and depict finer details and will be less prone to aliasing/moire.
So, it isn't just a pixel density issue. It is a resolution issue. And because FF sensors are larger, for the same pixel density, they have greater resolving ability.
That line of reasoning still works for me. But, of course, this is not my "area."
I think conropl is on the better track. The aliasing/moire is a frequency based issue. For the same image projected on the (same-size) sensor, a 50MP sensor has a different frequency at which a sampling error occurs than a 20MP sensor.
Ever watch a video of a car's wheel turning as the car drives down the street? As the car accelerates, the frame rate eventually gets caught by the wheel rotation rate (divided by the number of spokes or equivalent). As the wheel rate catches up to the frame rate, the wheel appears to be going backwards, then standing still, then forwards, then forwards faster still, then a blur, until the wheel frequency approaches 2x, etc. We all know the wheel isn't going backwards, but I'm sure the experienced videographers probably have a cheat sheet for frame rate vs. spoke count that shows the MPH speeds that are an issue. I think moire and pixel density is a similar thing (but certainly could be wrong).
Dave Throgmartin
03-30-2016, 11:52 PM
Nikon's used AA less full frame cameras with as few as 36 MP. I haven't seen anyone besides Fuji use AA less cameras with as few as 16 MP and they have a different color array.
That said, logic would say that moire/aliasing is strictly based off the pixel size given no AA filter.
Dave
Thought this might help to better understand aliasing issues with digital sampling methods. Moire patterns are what develop in a two dimensional array when the relative sizes of a repeating pattern in the subject compared to the sensor element size leads to aliasing. It can happen with any size of elements in a pixel grid.
http://users.wfu.edu/matthews/misc/DigPhotog/alias/