[biz-engineer]

I was curious If I could tell the visual difference between two A1 (~24x36") size paper prints, printed at 300ppi, of a photo captured with a Pentax K1 Mk II pixel shift and the same photo capture with a GFX100 normal exposure.
But since I do not have a GFX100 at hand, I decided to do a comparison by downloading the respective raw files (RAF and DNG) from DPReview lab comparison tool, both shot with prime lenses at f5.6 at ISO100 in lab conditions (tripod etc).

Then I processed each file the same way, as I would process them to make A1 size prints:
- Fired Silkypix Pro 10, set Raw conversion to Silkypix defaults for both files, default demosaic settings, default noise reduction, default sharpening, except I set the noise reduction mode to "fine details" for both the K1-2 and GFX100 files.
- Corrected both files for lens chromatic aberrations the same way, using Silkypix automatic correction tool
- Cropped the K1II image edges to 4:3 to match GFX100 aspect ratio
- Both files exported to default Silkypix exif compliant 97% Jpeg format, up-size 133% for the K1, down-size 75% for GFX100 to match printer resolution requirement of 300ppi of the photo lab for A1 prints.
- Loaded both Jpeg exports in Topaz Sharpen AI, using simple sharpening mode (the mode that's not using AI replacement for textures) and I let Topaz set the sharpening and noise reduction amounts automatically for each file.
- In each image, I then cropped a 3:2 rectangle of the most detailed area, containing small text, and lots of color detail (most relevant), and I printed both on A4 photographic paper at the photo lab.
- Took my +3 diopter close-up reading glasses to compare the prints at 6 or 7 inches viewing distance, best possible distance with the reading glasses

I drew an interesting conclusion.
What camera do you think resolves the best ?
Do you think the GFX100 worth $4000 more than a K1 mk II in pixel shift mode ?
How about in real life shooting conditions, with depth of field limitations , with or without a tripod ?

[surfar]

What camera do you think resolves the best ?

Well, the 144mp K1mk2.

When the GFX 1oo first appeared I saw a 1mp crop which looked very good BUT it was only on a 2mp screen

---------- Post added 09-12-22 at 08:02 PM ----------

How about

Comparing the GFX Pixel shift 400mp images to the K-1 PS, can you download them?

[rpjallan]

Is there any point to this? I think you have too much time on your hands...

[Fogel70]

In ideal conditions for both I do not think it will be much difference. It will probably be other factors that make the biggest influence in choosing one over the other.

If the price difference is worth it probably depend on several other things.
- Difference in available lenses.
- The scene may not always work well with pixel shift.
- Difference in workflow when using pixel shift. Where it takes longer to get confirmation of the end result.
- How much the price of the gear affect you.

[biz-engineer]

Is there any point to this? I think you have too much time on your hands...

There is a point. I print A1 as regular size. At that print size, I'm asking how much visible difference a GFX100 system makes for the money spent? It's a relevant test, at least for me.

---------- Post added 12-09-22 at 11:43 ----------

Well, the 144mp K1mk2.

Not quite.

Comparing the GFX Pixel shift 400mp images to the K-1 PS, can you download them?

I don't have a 400Mp GFX file, and it would be total overkill for A1 prints anyway.

---------- Post added 12-09-22 at 11:46 ----------

- Difference in available lenses.

That involves a trade-off. There a larger choice of FF lenses, especially for macro.

- Difference in workflow when using pixel shift. Where it takes longer to get confirmation of the end result.

Exactly.

In ideal conditions for both I do not think it will be much difference.

There is a small difference between the two , worth knowing / understanding for large prints.

[Terry C]

A member of a photographic FB group that I use, posts photos taken with GFX 100. Even with FB compression they always look super sharp. Far better than most.

[biz-engineer]

A member of a photographic FB group that I use, posts photos taken with GFX 100. Even with FB compression they always look super sharp. Far better than most.

It's the light, the post processing, and especially the sharpening at the final display size. I can tell the difference between 200 and 300ppi, but there is no way I can tell the difference between 300 and 400ppi, no matter how much effort. I even fool myself picking the wrong file / camera and I laugh when this happens.

[Roadboat24]

I doubt I would see any difference.

[pinholecam]

Largely irrelevant to me.
I print up to A3, beyond that, no space, no money, no patience (for the stressful workflow and test prints on same media that everything is correct before committing to print ).....

For my needs, mp largely stopped at 24mp which has been good enough for all my needs.
K1 at 36mp and the fantastic DR is the max that I willingly use.
The high mp sensors really don't interest me anymore and I'd only grudgingly buy one if they are the only ones available that I can use my lenses on.

[UncleVanya]

Well, the 144mp K1mk2.

Isn’t the k1 pixel shift only 36mp? Albeit with high color accuracy.

[biz-engineer]

Isn’t the k1 pixel shift only 36mp? Albeit with high color accuracy.

Yes and no. Color feature sampling pitch is 7.52um on the GFX100, to be corrected by approximately 0.79 if we want to compare to K1, that's 5.94um. K1 with pixel shift color sampling pitch is 4.89um.

[photoptimist]

Isn’t the k1 pixel shift only 36mp? Albeit with high color accuracy.

"It's complicated"

The Bayer color filter array on the K-1 and the GFX100 means that these cameras resolve different numbers of megapixels in the different color bands. With a Bayer filter, 1/4 of the pixels see the red channel, 1/4 of them see blue, and 1/2 of them see green.

Or to put it another way, 3/4 of the sensor is blind to details in red, 3/4 of the sensor is blind to to details in blue, and 1/2 the sensor is blind to to details in green. Putting a sharp lens in front of the Bayer color filter array sensor (that lacks an antialiasing filter) means that tiny details in red have a 75% chance of being lost!

In single shot mode, a K-1 resolves 9 Mpix in red, 9 MPix in blue, and 18 MPix in green.

In single shot mode, a GFX100 resolves 25.4 Mpix in red, 25.4 MPix in blue, and 50.8 MPix in green.

Pixel shift mode carefully shifts the sensor in a square pattern of 1-pixel increments so that every pixel in the image is sampled in red and blue and is sampled twice in green. This mode fully measures all the pixels in all the colors.

Thus in pixel shift mode, a K-1 resolves 36 Mpix in red, 36 MPix in blue, and 36 MPix in green (with better DR from average two samples).

For images with a lot of detail in the red or blue channel, K-1PS mode would beat GFX100. For images with a lot of detail in the green channel, the GFX100 would probably beat K-1PS mode.

[biz-engineer]

For images with a lot of detail in the red or blue channel, K-1PS mode would beat GFX100. For images with a lot of detail in the green channel, the GFX100 would probably beat K-1PS mode.

That's the difference I notice when comparing two prints , close look. The K1 print shows slightly crispier details in colorful area, but show slightly softer hard black on white edges than they are on the GFX100 print. So, overall, it's hard to tell which print look more detailed depending on where we look, especially when the eyes tend to look at colorful areas first. One thing however is that K1 pixel shift print still show some moiré patterns (on black & white details), which we don't see on the GFX100 print, so I conclude that the FA 77 limited @ f5.6 is still resolve contrast details smaller than the 4.89um pixel pitch can record, even with pixel shift enabled.

[UncleVanya]

Please note that y response was about the claim the pixel shit k-1 was a “144mp” image. I didn’t think it was accurate. Thanks for the clarifying info. I’m aware of the Bayer CFA and still think of the final result as a true color 36mp file since that’s precisely what gets created. Granted the explanations of the color detail loss without this are relevant and add to the discussion, in the end the file is “only” 36mp and has better color details than the non shifted file. Overall I like how this was discussed and dissected but I wanted to clarify the reason for my input.

[biz-engineer]

Granted the explanations of the color detail loss without this are relevant and add to the discussion, in the end the file is “only” 36mp and has better color details than the non shifted file. Overall I like how this was discussed and dissected but I wanted to clarify the reason for my input.

The important element to consider is also print size and visual acuity. At A1 print size, the GFX100 file is down-sampled (x 0.75) to match the 300ppi requirement of the print lab. The K1 file is up-sampled from 208ppi to 300ppi (x 1.33), with some pixels discarded at the edges to match the 4:3 ratio (advantage for GFX100). Human vision performance (my vision) isn't a linear function of feature size, sensitivity to detail drops quickly above 200ppi. And the paper print process also has its own (smoothing) response to small details, not has sharp as active RGB pixels from an LCD. That why I really wanted to emulate an A1 size prints (via A4 crop), from two cameras.