Moiré disaster on the E-5

Started Jul 10, 2013 | Discussions thread
John Sheehy
Forum ProPosts: 15,982
Like?
Re: Not nearly the same ...
In reply to Airmel, Jul 13, 2013

Airmel wrote:

Well as a matter of fact, what I said is quite different that what GB said. An anti-alias filter with a higher cutoff frequency is not "weaker", it simply utilizes a higher cutoff frequency.

An AA filter is usually two sheets of birefringent material, each of which takes a point of light, and turns it into two points of light.  They are layered at 90 degree angles from each other, to result in 4 points of light.  This is a dirty way to do it, but this is how it is done.The larger the spacing between the points, the stronger the AA filter is said to be.  Even if we are talking about the spacing of the four points relative to pixel spacing, it is still true that a weaker AA filter (as measured in pixels) is needed for higher pixel densities, because diffraction and other optical issues are going to attenuate higher frequencies, anyway.  A sensor with large pixels has to deal with the possibility of a very sharp lens with very little diffraction (say, a Sigma 35mm f/1.4 art lens at f/2, or wherever its sweet spot is).

The required cutoff frequency is well defined by the sampling theorem. It is directly related to the sampling frequency and is nowhere near infinite.

I didn't say that it was infinite in any actual AA filter.  I said that one that was designed to be infinite would simply not exist at all.  I said that because your language seemed to suggest to someone less in tune with things that there was some special achievement in designing an AA filter with a "high cut-off frequency", when the reality is that a high-cutoff is, in actuality, almost no filtration at all.

The more pixels per unit distance, the higher the cutoff frequency of the anti-alias filter will need to be. Very simple really.

Now as a practical matter, it turns out that virtually all optical anti-alias filters are very poor performers. The technology simply doesn't exist to build a cheap optical low pass filter which has anywhere near the performance of its electronic signal processing counterpart. For this reason you might rightly refer to all optical anti-alias filters as "weak".

AFAIK, they all turn one point into 4, so strength/weakness can on refer, in practical concerns, to the spacing of the 4 points, either absolutely or relatively.

However so, there is no validity to the notion that an anti-alias filter with a higher cutoff frequency is "weaker" in some way. If you adjusted your sports car's speed limiter from 125 MPH to 155 MPH, you wouldn't say the speed limiter is now "weaker". It would simply be tuned to a different speed limit. The speed limiter's performance above and below the new limit would be the same as that seen above and below the previous limit.

Active analogies for passive phenomena?

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
JohnNew
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow