[jpzk]

Just saw this:
Sony's sensor roadmap includes a 150MP medium-format chip for 2018: Digital Photography Review

What will ne next ... ? 200+ MP chips ?
Storage nightmare (for me) this would be !

[Mistral75]

What might be more interesting for us Pentaxians is 100MP 33x44 BSI CMOS sensor (IMX461) said to be available in 2018... when the successor to 645Z is supposed to be launched.

Do you remember kenspo being in trouble early last year because he wrote that Ricoh Imaging and Sony had cut a deal on a 100MP sensor?

[pathdoc]

What will ne next ... ? 200+ MP chips ?
Storage nightmare (for me) this would be !

My fun conspiracy theory is that the digital camera sensor manufacturers are in cahoots with the manufacturers of computer equipment, particularly external (and for desktops, internal) hard drives.

[D1N0]

You can Buy 10TB hard disks already. By the time Average Joe kan afford to buy a large sensor system they will be larger still. No problem.

[ffking]

You can Buy 10TB hard disks already. By the time Average Joe kan afford to buy a large sensor system they will be larger still. No problem.

True - post processing might prove more of a problem without a considerable upgrade to the PC

[VoiceOfReason]

35tb of space on my main computer, 30tb in my NAS, and unlimited cloud backup, so space is no problem, processing isn't either with a 6 core i7 and 64gb ddr4 ram. What is the problem is the cost of medium format vs my pay check.

[RGlasel]

What will ne next ... ? 200+ MP chips ?

I noticed that Sony's proposed applications for these 51 x 41mm sensors aren't conventional cameras. Heat dissipation and power requirements limit their practicality for handheld cameras and the benefit of the increased resolution is only with enlargement of the entire image (as opposed to zooming in on a smaller area by changing focal length). Even with existing digital MF cameras, higher resolution doesn't translate into better images by itself, it also takes the optical qualities of the lenses and perceived depth of field to produce photographs that look different than ones taken with smaller format cameras.

[biz-engineer]

150 MP medium-format chip for 2018 ??

The K1 was already at 144Mpixels in 2016, yes (Pixel shift) 4 x 36 = 144.

[alpheios]

The K1 was already at 144Mpixels in 2016, yes (Pixel shift) 4 x 36 = 144.

Is that so? The way I understand how Pixel Shift works is that it still produces a 36,4 megapixel image but each pixel contains vastly more color information and thus making it comparable to a higher resolution image while still having the same effective pixel count.

[biz-engineer]

Is that so? The way I understand how Pixel Shift works is that it still produces a 36,4 megapixel image but each pixel contains vastly more color information and thus making it comparable to a higher resolution image while still having the same effective pixel count.

In the end, all terminals (Plasma, LCD, OLED , and printed) treat information and display color according to juxtaposed RGB pixels (not stacked).
So, when pixel shifted stacks are converted to JPEG, the image stack is actually downsampled from 144Mp RGB back to 36Mp RGB. But the RAW stack could be converted as is, into 144Mp JPEG, it wouldn't resolve true 144Mp anyway, but the pixel count would be there.

[alpheios]

In the end, all terminals (Plasma, LCD, OLED , and printed) treat information and display color according to juxtaposed RGB pixels (not stacked).
So, when pixel shifted stacks are converted to JPEG, the image stack is actually downsampled from 144Mp RGB back to 36Mp RGB. But the RAW stack could be converted as is, into 144Mp JPEG, it wouldn't resolve true 144Mp anyway, but the pixel count would be there.

Thanks for clearing that up. I get it now.

[biz-engineer]

each pixel contains vastly more color information

Vastly? The difference between stacked and bayer interpolation is something like 10%. The difference is more noticeable from the noise reduction of stacking.
For a wider range of colors, pixel shift should be combined with a different color filter array having wider range. The color filter array bandwidth is defined as a tradeoff between pixel S/N noise (limited bandwidth of the photodiode that lies behind the color filter) and the range of colors that can be reproduced. If the red of the red color filter is shifted more in the direction of IR wavelength, the ability for the camera to reproduce real colors is enhanced, but the S/N and dynamic range of the red component decreases. On the other hand, if the red of the red color filter is shifted towards shorter wavelength, the pixel S/N and dynamic range are improved but the ability to code real colors is reduced. No color can be reproduced outside of the color gamut ( https://en.wikipedia.org/wiki/Gamut ). That triangle is defined by the color filter array and can't be changed by pixel shift.

[alpheios]

Thanks a lot for your posts biz-engineer. That's why I love pentaxforums. There are so many people with so much knowledge here to learn from.

[rechmbrs]

In the end, all terminals (Plasma, LCD, OLED , and printed) treat information and display color according to juxtaposed RGB pixels (not stacked).
So, when pixel shifted stacks are converted to JPEG, the image stack is actually downsampled from 144Mp RGB back to 36Mp RGB. But the RAW stack could be converted as is, into 144Mp JPEG, it wouldn't resolve true 144Mp anyway, but the pixel count would be there.

Pixel Shift (PS) is not a stacking process. PS is a compositing process where the missing color values in the Bayer pattern come from the other shots. Therefore the PS output has 3 recorded (real) color values at every pixel location rather than having to interpolate 2/3 of the color values as the standard Bayer pattern requires. For the K 1, there are 36 million pixels with or without PS. Or 36 million color values versus 108 million for PS.
Forgetting movement between shots, the final PS image has more consistent s/n. The Bayer pattern interpolation image has noise induced because interpolation does not match what should be for those color values.
There a number of threads where this discussed. If output of the raw converter is to a non-lossy compression format, there will be 108 million color values for 36 million pixel locations. Jpg is lossy but there will be the same number but the compression distorts them.

RONC

[biz-engineer]

Or 36 million color values versus 108 million for PS.

Ok, how much is the color accuracy improvement of PS vs interpolation vs Adobe RGB ?
I have the impression that using Adobe RGB instead of sRGB brings way more realistic color coding then anything pixel shift. Perhaps a combination of both is the best, under the assumption that the color space of the filter array is at least as large as Adobe RGB color space.