PDA

View Full Version : Anti alias filter...why?


crimbo
4th April 2011, 06:54 PM
This is from my 620 with the IR and AA filter ripped out...
So what does the AA filter really do...often?
How can I check the resolution of this system?

http://www.paddle.shetland.co.uk/I3253280.jpg

http://www.paddle.shetland.co.uk/I3283498.jpg

http://www.paddle.shetland.co.uk/I3273430c.jpg

Garrie
4th April 2011, 07:34 PM
I think its something to do with moire and reducing the effects of it.

crimbo
4th April 2011, 07:58 PM
I think its something to do with moire and reducing the effects of it.

but I havent seen any...yet...I think

Nick Temple-Fry
4th April 2011, 10:23 PM
Well you should only see Moire when at close to the resolution limit of the sensor - ie when the sampling frequency (in our case pixel seperation) is less than N times the frequency of the pattern being sampled, in the best case for photography that is a minimum of 2 pixels, though in reality it is probably three or more.

When an item of repeating detail has a frequency higher than this then the digital conversion may see it as a combination of more than one lower frequency. Giving the moire effect of a coarser pattern overlaying an area which we know contains finer details.

This is all explained in the Nyquist-shannon theorem (if you are one of those who can read the necessary maths).

The Anti-Alias filter prevents the frequency of detail exceeding the N times the pixel spacing, it therefore sets the upper resolution level for the sensor combination.

Now if you take the AA filter away then the resolution of the sensor should approach its theoretical value (where the smallest repeating pattern that can be resolved should be between 2 or 3 times the pixel seperation).

But you also need take into account diffraction limiting the resolution of the lens (about which Ian posted an article quite recently).

Actually I suspect it's all 'even harder' as our sampling pattern is different for Red Green Blue and will differ again depending on the relationship between the alignment of the pixels and that of the pattern.

None of the provided images seem to have much in the way of information that would generate moire (repeating high frequency patterns). So AA/moire is really quite difficult and unpredictable, but you wouldn't expect to see it without a lot finer detail than your pictures show. Though it may show up in a crop of a fabric surface in a portrait.

Nick

theMusicMan
5th April 2011, 06:55 AM
Nick; I think I speak for everyone here... you are, quite simply, a technical God!

What a post, simply brilliant. Thanks for explaining.

Ian
5th April 2011, 09:43 AM
With no IR filter, your sensor is recording a much wider spectrum into the longer wavelengths. This is light that is not visible to the human eye. It has the effect of lowering contrast and skewing colours, and I think both these effects are demonstrated in your shots.

Ian

Howi
5th April 2011, 10:26 AM
With no IR filter, your sensor is recording a much wider spectrum into the longer wavelengths. This is light that is not visible to the human eye. It has the effect of lowering contrast and skewing colours, and I think both these effects are demonstrated in your shots.

Ian

Also check out the Leica M8 reviews ............

crimbo
5th April 2011, 04:30 PM
thanks folks...not bothered about the colours as the camera is mainly for astro work or an IR pass filter as in the first image...
I am really wondering how much moire really turns up and do the cameras really need an A-A filter

Also I have a 50 f2 so how could i work out the resolution of this system compared to a normal 620 ?

Nick Temple-Fry
5th April 2011, 05:01 PM
I do wonder whether the fact that the camera has been adapted so as to be sensitive to IR has reduced the resolving power of the system sufficiently to mean that moire is unlikely.

OK, as I've previously suggested removing the AA filter allows the sensor to perform closer to its theoretical limit (for visible light anyway). IR has a different refractive index because of its different wavelength to that of visible light (relationship known as dispersion). Conventional lenses are designed to take light in the visible range and bring it to focus in the same physical plane parrallel to the lens, stick in a wavelength outside of the design parameters for visible light and the only thing that can be predicted easily is that it probably wont work as well.

So maybe you wont see Moire because the lens resolving power has now fallen below that of the sensors.

Still, you will probably have fun trying it out on a repeating pattern at varying distances/magnifications. Though I suspect some of the more techy physics/astronomy forums have covered this far more proficiently than I can.

Nick

crimbo
5th April 2011, 06:51 PM
yer right John...nowt seen on the forums

nick - Without the IR filter the wider range of wavelengths will make it more difficult for the lens to focus them all at the same time so yes I understand that overall resolving power may have been reduced as the Airy disk is larger...and with an IR pass filter the lens will not be optimally focussing and hence resolve less...but must find a pattern to practice with...net curtain!