Yes.
Would it be commercially viable? Absolutely not.
Would it ruin DR because you're adding at least one to two orders of magnitude more circuitry and read noise? Yes, and that's exactly what's happening to stacked sensors and global shutter.

Wait a decade or two. SPAD being potentially more likely.

>>4523318
I can't think of a single useful application for that, but technically it'd be rather easy to implement and it wouldn't ruin DR either unlike the other anon said

>>4523322
>I can't think of a single useful application for that
For still photography, there's at least HDR capture without temporal bracketing. Spatially varying exposure gets you extended scene DR in a single frame.

>it'd be rather easy to implement
Define rather easy.
Starting from a global sensor, per-pixel exposure would be "rather easy" but it's still more circuitry and more computing per individual pixel.
For per-pixel gain, PGA and ADC reference are column-parallel by architecture, so the only real in-pixel gain knob is a switchable conversion-gain capacitor giving you two or three discrete states (double or triple conversion gain architectures). A continuously programmable per-pixel amplifier is not "rather easy".

>it wouldn't ruin DR
Let's say we make this an evolution of global shutter sensors. Because you pretty much need a global shutter architecture for this anyway:

The storage node and its extra transistors eat area on top of the per-pixel control logic you want to add. Two area budgets competing for the same silicon. More switching near the diode means more clock feedthrough and coupling, and active logic adjacent to the photodiode. This means more heat, more dark current, therefore more read noise. Dynamic range is full-well over read noise, so a raised floor is a direct DR loss.

Current global shutter charge-domain photosites have a smaller full-well capacity than voltage-domain ones on non-GS cameras, which means lower DR in any case for now.

>>4523318
Yes, it is called CID sensor vs the global gain design of CCD and CMOS. Most laboratory spectrometers use it.

>>4523327
Modern CID sensors are CMOS, only CCD have global gain.
Still no per-pixel gain control on CID doughbeit.

>>4523326
>Spatially varying exposure gets you extended scene DR in a single frame.
You'd still need to take two photos, one to find out how to correctly expose the second one. Between the photos there'd be a break of tens of milliseconds for computing and programming. At this point you can just do regular exposure bracketing.

>For per-pixel gain, PGA and ADC reference are column-parallel by architecture, so the only real in-pixel gain knob is a switchable conversion-gain capacitor giving you two or three discrete states (double or triple conversion gain architectures). A continuously programmable per-pixel amplifier is not "rather easy".
Varying the gain is an antiquated approach anyways, but it'd absolutely be doable to chang it between each pixel given enough readout time budget. You'd generate the voltages with a low number of fast DACs and then 'copy' the voltage output with local sample/hold buffers.

>Because you pretty much need a global shutter architecture for this anyway
No it would be much easier to start with a conventional rolling shutter sensor.

>in 20 years your camera will analyze the landscape you're shooting, it will use ChatGPT to understand where the sky is and will manually lower the gain on the sensor's pixels, simulatimg a $2 graduated filter from 1955

>>4523465
There are already graduated filter modules available in the cameras.

>>4523465
No thanks.

>>4523431
>No it would be much easier to start with a conventional rolling shutter sensor.
It would not. With a GS you only need a pixel latch and an additionnal transfer gate for individual exposure control. Sequential row-level exposure control from rolling shutter is further away to be adapted into pixel-level exposure control as easily

>>4523318
I asked this question many years ago and it was pointed out to me that the magic lantern custom canon firmware had the ability to control the ISO on a row by row basis and they were able to use that ability to implement the dual iso feature where alternating rows of sensitivity are interleaved.

Funny how cameras are all about ADCs. In my work I design circuits using ADC chips. But I think those things are very much unlike the ones in cameras.

iso is just the exposure slider in photoshop, shutter speed is in camera, possible but lots of rnd

>>4523694
how is it funny? it's the literal single point of measurement of the actual light of the scene, of course it's the most crucial part

>>4523753
>iso is just the exposure slider in photoshop
holy retard

>>4523765
With more and more linear ISO invariance sensors it pretty much is, you might get a second base ISO at 400 or 600 where thenoise reduction hardware kicks in but pretty much any high ISO shot is just this base ISO digitally clipped and exposure increased, very similar to increasing the exposure slider. The noise will be the same, all exposure properties will be the same. It has been for 10 years.

don't know where else to ask this but you lot seem savvy enough

What is the easiest way to have a GPS app running on my phone while shooting photos, and then I dump the photos and the geolocation data to my computer and have the nearest/interpolated location attached to the EXIF of each photo?

>>4523318
>>>4523318
>control ISO and shutter speed (if electronic) on a per pixel basis
Why not do in post process on the raw pixels? Cannot see a different outcome.

>>4523778
Dual conversion gain sensors are not noise reduction hardware you dunning kruger ass schizo retard snoyim

Its basically turning off a high dynamic range circuit that stores more charge to switch to a more sensitive and less noisy one that cant store as much charge but detects finer increments.

Noise reduction is what pentax does. No extra light is stored. No extra shadows are detected. Its just a noise reduction slider on a chip for when the camera processor is too weak to run a good NR algorithm.

Modern sensors read both of these circuits simultaneously (sony a7v, lumix s1rii, om om1ii) or sequentially (lumix g9ii).

File: convert.png (75 KB, 2560x3300)

75 KB PNG

>>4523318
Apparently that's what's described in this Nikon patent
>US-20260281583-A1
But I can't read legalese so correct me if I'm wrong. Could just be describing a faster sensor for all I know.

>An image capturing element includes a first substrate having a plurality of pixel blocks including at least one pixel, the plurality of pixel blocks being arranged to be aligned in a row direction and a column direction, and a second substrate having a plurality of control blocks including a conversion unit which converts a signal output from the pixel into a digital signal and a through electrode unit configured to output the signal converted into the digital signal by the conversion unit, the plurality of control blocks being arranged to be aligned in the row direction and the column direction.