Since the A1 lacks a low-pass filter (OLPF), it will oversample a high-frequency image projected onto the sensor and produce artifacts. The classic moiré pattern on fabric is an example, and the reason I usually shoot around f/20 for studio tabletop. Reducing the aperture size causes the Airy disk to enlarge and therefore reduce the image resolution. This is an effective low-pass filter and restores true image capture (no aliasing).
Below is text at a point size close to sensor resolution. First is f/2.8 aperture:
Second is f/16:
2nd image has better legibility due to lack of artifacts, although we are (deliberately) near the limit of sensor resolution. Both images had curves applied in post, and 2nd had sharpening to match apparent sharpness of the 1st. Here is the uncropped scene showing the document lit by manual flash as taken by Sigma 45C, a remarkably sharp prime if critical focus can be attained (I used contrast AF):
In summary, if you like false data and crunchy images then OLPF is nothing for you to miss.
I have mixed feelings about the removal of the low-pass filter. A few years ago, I was trying to decide between the Pentax K-5 II and the K-5 IIs versions. The s version had the low-pass filter removed for sharper images at the pixel level, which was the only difference between the 2 cameras. And there were a lot of discussions at the time about these two options. I finally opted for the non-s version (the one with a low-pass filter) because, as much as I love pixel-peeping, I felt that the s version was damaging the tonality and creating moiré in many instances.
A few years later, Pentax found a clever solution in later bodies, and one that is alas not replicated by Sony: remove the low-pass filter but give the option to simulate it by sending microscopic vibrations to the sensor.
Manu-K1 wrote:
... because, as much as I love pixel-peeping, I felt that the s version was damaging the tonality and creating moiré in many instances.
A few years later, Pentax found a clever solution in later bodies, and one that is alas not replicated by Sony: remove the low-pass filter but give the option to simulate it by sending microscopic vibrations to the sensor.
Pentax was being super clever. That's a valid solution to oversampling, although I could imagine motion artifacts. The introduction of pixel-scale noise can be done in post—no need to commit the sensor to frequencies that exceeds its Nyquist frequency. The same noise in film comes from visible grain and gives the sensation of detail: appreciated by the human eye except in the sky. Nonetheless, image noise and unsharp masking is nothing but eye candy, and cannot create valid image detail from thin air.
I should add that the Sony demosaic'ing machinery for the A1 is outstanding, and the best I've seen in any digicam. There is virtually no color aliasing at pixel scale in the oversampled image. Pixel shifting is usually required to achieve these results. The images above were demosaic'ed in-camera and stored in HEIF 4:2:2 format.
snapsy wrote:
Have you ever compared slightly defocusing the lens vs shooting at a small aperture?
Yes, but only as an experiment. It's practically impossible to shift the plane of focus to the precise degree needed. Except for flat art, the plane of focus is shifted backward (or forward) and something else is now being oversampled. Plus, it looks like you missed focus.
rico wrote:
Yes, but only as an experiment. It's practically impossible to shift the plane of focus to the precise degree needed. Except for flat art, the plane of focus is shifted backward (or forward) and something else is now being oversampled. Plus, it looks like you missed focus.
Another idea which occurs to me...shooting the same scene at both the optimal aperture (for desired DOF) and at the smallest aperture (induce diffraction to reduce aliasing), then blending the two exposures together selectively in post, using either a region-oriented strategy (for your photo, that would be using the small aperture exposure for text and optimal aperture for rest), or some type of algorithmic merge based on the sampled frequencies (ie, there are algorithms which can reliably detect aliasing - the output of those could feed into a merging algorithm).
When the Nikon D800E (no low-pass filter) was released 14 years ago, I purchased it. I often shoot with two bodies, so I later acquired a D800 (with LP filter) and shot side-by-side with the D800E.
Later, I went through the same thing with Canon. I purchased the 5DSR (no LP filter) and then, later, a 5DS (with LP).
After shooting many thousands of images with these camera's and comparing images, I found the lack of LP filter to be highlly advantageous in terms of sharpness. I seldom had problems with moire. I am now biased to where I will not buy a camera with a LP filter. The exception is my RX1RII, which has a variable LP filter (which I turned-off).
rico wrote:
Since the A1 lacks a low-pass filter (OLPF), it will oversample a [...](I used contrast AF):
[image]
In summary, if you like false data and crunchy images then OLPF is nothing for you to miss.
I have this simple question:What did you do for using contrast AF? AF-S only or something else?
Have you tried with the pixel shifter? I am just fine without the AA filter on high-res bodies like the a7rV and a7rVI. It seems that the AI NR likes sharp pixels and plenty of them.
