fredmiranda.com Login
This topic is in the 2021 archive. You can read it and like its posts here. It is read only.

Archive 2021A1 AA filter

  
 
cgarcia
Offline

p.1 #1A1 AA filter
In case the A1 (at the left) finally has an AA filter, it is much weaker compared to the one on the R5 (at the right).

(Source: https://www.dpreview.com/articles/8956705170/sony-a1-added-to-studio-test-scene )

https://www.etherpilot.com/photo/test/misc/aliasing__A1_vs_R5.png
Feb 16, 2021 at 01:39 PM
Stoffer
Offline
Details
p.1 #2A1 AA filter
Yeah, it doesn’t look like the A1 have an AA filter to my eye but it is difficult to compare those two because of different lenses and probably different distance to the subject. Moire will of course change based on that.
Feb 16, 2021 at 01:49 PM
Daran
Offline
Details
p.1 #3A1 AA filter
The A1 has Pixel Shift, which would be impossible with an AA filter.
Feb 16, 2021 at 02:02 PM
cgarcia
Offline
Details
p.1 #4A1 AA filter
Daran wrote:
The A1 has Pixel Shift, which would be impossible with an AA filter.


Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )
Feb 16, 2021 at 02:29 PM
RAF09
Offline
Details
p.1 #5A1 AA filter
I believe the a1 does not have an AA filter.
Feb 16, 2021 at 02:40 PM
Daran
Offline
Details
p.1 #6A1 AA filter
cgarcia wrote:
Daran wroteThe A1 has Pixel Shift, which would be impossible with an AA filter.
Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )

OK, fair enough, not impossible, just pretty inefficient. With an AA filter you are already oversampling, basically every pixel being the sum of the local area. The A1 uses 16 images for a full Pixel Shift scan, so full RGB per pixel times 4. With an AA filter that data would be extremely redundant.
Feb 16, 2021 at 02:44 PM
Holger
Offline
Details
p.1 #7A1 AA filter
Daran wrote:
cgarcia wrote
Daran wroteThe A1 has Pixel Shift, which would be impossible with an AA filter.
Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )
OK, fair enough, not impossible, just pretty inefficient. With an AA filter you are already oversampling, basically every pixel being the sum of the local area. The A1 uses 16 images for a full Pixel Shift scan, so full RGB per pixel times 4. With an AA filter that data would be extremely redundant.


I understand it differently. You are not oversampling, you are eliminating/damping small scales close to Nyquist, not sampling at higher frequency than Nyquist.

The Canon R5 uses a "low-pass filter that reduces false color and moire is composed of two independent plates, a birefringent crystal that vibrates ultrasonically and an infrared absorbing glass."
Some think it will be the same one as in the 1dxiii (16 point filter).
Feb 16, 2021 at 03:01 PM
Daran
Offline
Details
p.1 #8A1 AA filter
Holger wrote:
Daran wrote
cgarcia wrote
Daran wroteThe A1 has Pixel Shift, which would be impossible with an AA filter.
Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )
OK, fair enough, not impossible, just pretty inefficient. With an AA filter you are already oversampling, basically every pixel being the sum of the local area. The A1 uses 16 images for a full Pixel Shift scan, so full RGB per pixel times 4. With an AA filter that data would be extremely redundant.
I understand it differently. You are not oversampling, you are eliminating/damping small scales close to Nyquist, not sampling at higher frequency than Nyquist.

With a good lens at its best aperture even high resolution sensors are undersampling. Meaning the input can change faster than it is sampled. Increasing the sampling frequency yields more non-redundant data.

An AA filter deliberately limits the optical resolution (aka input frequency). So now the the sensor oversamples (same sampling frequency works on a lower input frequency). Which is a good thing when you want to avoid Moiré. It still means high frequency inputs are filtered by the low pass. Since you are already oversampled, increasing the sampling frequency has only limited benefits.

Of course this isn't a black and white affair (pun unintended). The weaker the low pass, the more beneficial is increasing the sampling.

The Canon R5 uses a "low-pass filter that reduces false color and moire is composed of two independent plates, a birefringent crystal that vibrates ultrasonically and an infrared absorbing glass."
Some think it will be the same one as in the 1dxiii (16 point filter).

Maybe. Cute stuff. Still a low pass filter.

What would be a huge improvement would be to selectively low pass only relative to the Bayer filters. So only the red and green part of a blue pixel would be spread to the red/green neighbors, not creating redundancies. Could reduce noise by up to 70%. Would also noticeably reduce Moiré (though not to the level of an AA filter). I don't expect the manufacturing of such a filter to be possible anytime soon, though. If even physically possible.
Feb 17, 2021 at 03:31 AM
Holger
Offline
Details
p.1 #9A1 AA filter
Daran wrote:
Holger wrote
Daran wrote
cgarcia wrote
Daran wroteThe A1 has Pixel Shift, which would be impossible with an AA filter.
Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )
OK, fair enough, not impossible, just pretty inefficient. With an AA filter you are already oversampling, basically every pixel being the sum of the local area. The A1 uses 16 images for a full Pixel Shift scan, so full RGB per pixel times 4. With an AA filter that data would be extremely redundant.
I understand it differently. You are not oversampling, you are eliminating/damping small scales close to Nyquist, not sampling at higher frequency than Nyquist.
With a good lens at its best aperture even high resolution sensors are undersampling. Meaning the input can change faster than it is sampled. Increasing the sampling frequency yields more non-redundant data.

An AA filter deliberately limits the optical resolution (aka input frequency). So now the the sensor oversamples (same sampling frequency works on a lower input frequency). Which is a good thing when you want to avoid Moiré. It still means high frequency inputs are filtered by the low pass. Since you are already oversampled, increasing the sampling frequency has only limited benefits.

Of course this isn't a black and white
...Show more →
Earlier quoteThe Canon R5 uses a "low-pass filter that reduces false color and moire is composed of two independent plates, a birefringent crystal that vibrates ultrasonically and an infrared absorbing glass."
Some think it will be the same one as in the 1dxiii (16 point filter).


There is no perfect cutoff in spectral space, filter strength gradually decreases.
A nice take on the Canon filter is found here:
https://www.strollswithmydog.com/canon-high-res-gd-lpf-aa/
Feb 17, 2021 at 04:35 AM
Daran
Offline
Details
p.1 #10A1 AA filter
Holger wrote:
Daran wrote
Holger wrote
Daran wrote
cgarcia wrote
Daran wroteThe A1 has Pixel Shift, which would be impossible with an AA filter.
Why?

In fact, there are cameras with such feature and AA filter (see https://www.dpreview.com/forums/post/64836280 )
OK, fair enough, not impossible, just pretty inefficient. With an AA filter you are already oversampling, basically every pixel being the sum of the local area. The A1 uses 16 images for a full Pixel Shift scan, so full RGB per pixel times 4. With an AA filter that data would be extremely redundant.
I understand it differently. You are not oversampling, you are eliminating/damping small scales close to Nyquist, not sampling at higher frequency than Nyquist.
With a good lens at its best aperture even high resolution sensors are undersampling. Meaning the input can change faster than it is sampled. Increasing the sampling frequency yields more non-redundant data.

An AA filter deliberately limits the optical resolution (aka input frequency). So now the the sensor oversamples (same sampling frequency works on a lower input frequency). Which is a good thing when you want to avoid Moiré. It still means high frequency inputs are filtered by the low pass. Since you are already oversampled, increasing the sampling frequency has only limited benefits.

Of course this isn't a black and white affair (pun unintended). The weaker the low pass, the more beneficial is increasing the sampling.

Earlier quoteThe Canon R5 uses a "low-pass filter that reduces false color and moire is composed of two independent plates, a birefringent crystal that vibrates ultrasonically and an infrared absorbing glass."
Some think it will be the same one as in the 1dxiii (16 point filter).
Maybe. Cute stuff. Still a low pass filter.

What would be a huge improvement would be to selectively low pass only relative to the Bayer filters. So only the red and green part of a blue pixel would be spread to the red/green neighbors, not creating redundancies. Could reduce noise by up to 70%. Would also noticeably reduce Moiré (though not to the level of an AA filter). I don't expect the manufacturing of such a filter to be possible anytime soon, though. If even physically possible.
There is no perfect cutoff in spectral space, filter strength gradually decreases.

Sure.

A nice take on the Canon filter is found here:
https://www.strollswithmydog.com/canon-high-res-gd-lpf-aa/

Interesting stuff. His model does seem to ignore the effect of the Bayer filter, though. Therefore I don't expect the real effect is nearly that neat. Has someone measured some hard data?
Feb 17, 2021 at 06:03 AM
f1atlux
Offline
Details
p.1 #11A1 AA filter
Sony Korea's A1 unpacking show has just ended.. In QnA session, Show host said that A1 has an AA filter. Hope this helped.


This is the link to the source and time he mentioned this was at 2:19:48 but it's language is Korean and subtitle is not provided.
Feb 17, 2021 at 08:43 AM
dclark
Offline
Details
p.1 #12A1 AA filter
Sony's pixel shift has two modes. The first involves 4 one pixel shifts of the sensor so that each pixel samples RGBG. Each pixel has full color data so that the need to demosaic the image is eliminated. If the color of the pixels is mixed by an AA filter this process is not possible, or at least greatly reduced in effectiveness.

A lot of the internet chatter about AA filters on the A1 is due to errors in posting Sony specs for the A1. They were not taken from Sony's published specs. Sony's published specs at this time are silent on the issue of the AA filter (and some other important issues).

There is no AA filter on the A1.

I will offer the following bet. If I am wrong I will post a 4K video showing me eating a crow, feathers and all. But only if someone will take the other side of this bet. I will provide the crow.

Dave
Feb 17, 2021 at 09:32 AM
Holger
Offline
Details
p.1 #13A1 AA filter
dclark wrote:
Sony's pixel shift has two modes. The first involves 4 one pixel shifts of the sensor so that each pixel samples RGBG. Each pixel has full color data so that the need to demosaic the image is eliminated. If the color of the pixels is mixed by an AA filter this process is not possible, or at least greatly reduced in effectiveness.

A lot of the internet chatter about AA filters on the A1 is due to errors in posting Sony specs for the A1. They were not taken from Sony's published specs. Sony's published specs at this time are
...Show more →

I think it unlikely too, given the images seen so far.
Feb 17, 2021 at 09:34 AM
Douglas L
Offline
Details
p.1 #14A1 AA filter
Someone posted in DPR that Image Resources just got confirmation from Sony that the A1 does not have an AA filter.
Feb 17, 2021 at 01:37 PM
dclark
Offline
Details
p.1 #15A1 AA filter
Douglas Liu wrote:
....just got confirmation from Sony that the A1 does not have an AA filter.


The crows that hang around my yard are relieved to hear it.
Feb 17, 2021 at 02:30 PM
FM ForumsSony ForumJoin Upload & Sell
  
 

You are not logged inLog inRegister