[redcat]

Magic Lantern can do with Canon cameras: Allow various pixels on the sensor to use different ISO, thus making a single image with much higher DR, but smaller MP. So you have 36MP with regular DR, or you can use this function to get lower MP, but much higher DR.

wow, this really exists ?!? sounds amazing, if someone can tweak the Pentax firmware to do the same @_@

[Na Horuk]

wow, this really exists ?!? sounds amazing, if someone can tweak the Pentax firmware to do the same @_@

Found a thread from 2013, so things probably improved since then.
Dual ISO - massive dynamic range improvement (dual_iso.mo)

Yess, it looks promising to those of us that don't really need a lot of MP, but would enjoy a boost in overall IQ

[patarok]

_DR_
.

[jsherman999]

If the DR increase as the pixel size decreases, then point & shots would have a higher DR then FF.

Phone cameras have a small DR range because the sensors are tiny, and they're collecting much less total light for the same exposure. There is no (or very little) causal link to pixel size there.

DR vs. MP is a false choice, largely, when we're comparing apple-apples and staying in the same sensor tech. There can be some correlation based on product life cycle, but we shouldn't take a manufacturers efforts at any point in time as a definitive statement about physics - Sony wants to sell a lot of lower-MP DR monsters, then provide the higher-MP replacements - with the same amount of DR or better - as 'upgrade' bait to those same folks, or to the folks who care more about resolution and want a reason (higher DR) to upgrade from their D800, etc.

And there have been pages upon pages and pages (read link for example) written about this, with everyone in the industry who bothers to read and contribute to dpreview chiming in. High image DR is not tied directly to 'larger pixels'.

"...With respect to why a high MP camera might have better DR than a lower MP one, it has to do with the question of the conversion gain, above. The conversion gain controls the correspondence between apparent photon noise and electronic noise. Imagine ten photoelectron corresponds to 1 millivolt at the pixel output. That means that one millivolt of electronic noise will look like one photon's worth of noise. If we now double the conversion gain so that 5 photoelectrons gives 1 millivolt, then that same millivolt of noise looks like 5 photons worth of noise rather than 10, we have effectively halved the read noise. Now, above I explained that the conversion gain needs to be designed proportional to the pixel capacity. So, imagine that we halve the pixel size, which means it gathers half the number of photoelectrons at the same exposure, then we can double the conversion gain resulting in half the read noise per pixel. What happens when we look at the noise of the two sub-pixels together? Noise adds 'in quadrature' or 'root-mean-square'. So, the combined noise of the two pixels is not double the noise in each, but just 1.4 (square root of 2) times. The signal is the same, so viewed at the same scale as the original sensor, we've increased the DR by 1.4 times." ~ Bob Newman

[Ratcheteer]

High image DR is not tied directly to 'larger pixels'.

correct, something still has to interpret the signal from the pixels (adc, processor, software etc) the end result is truncated

Originally posted by Na Horuk
Magic Lantern can do with Canon cameras: Allow various pixels on the sensor to use different ISO, thus making a single image with much higher DR, but smaller MP. So you have 36MP with regular DR, or you can use this function to get lower MP, but much higher DR.

its not various pixels its alternating lines of pixels sampled at different sensitivities (as a ex 5d mark2 magic lantern user i only really found this usable in 2 situations, places where your not allowed flash, and shaded areas with strong back lighting) most people using magic lantern are using it for raw video not dual iso

[jsherman999]

From this recent dpreview article on the Sony A7 RII (LINK)

"...In the graph above, made by our own resident forum expert Bill Claff, the a7R II maintains higher dynamic range than the a7S at all ISOs. This is particularly impressive given its smaller pixels. The advantage at low ISO, we think, comes from the fact that it's easier for a 14-bit ADC (or 13-bit, according to experts) to quantize the smaller signals of smaller pixels than the larger signals of larger pixels, which need more bit-depth to properly quantize both the bright and dark signals generated by large pixels. How's that for overturning the myth that more pixels always mean more noise?"

.

[Ratcheteer]

The advantage at low ISO, we think, comes from the fact that it's easier for a 14-bit ADC (or 13-bit, according to experts) to quantize the smaller signals of smaller pixels than the larger signals of larger pixels, which need more bit-depth to properly quantize both the bright and dark signals generated by large pixels.

i like how he states "we think" and how he used the word quantize (this reinforces my point of view that processing is just as important when it comes to DR)

How's that for overturning the myth that more pixels always mean more noise?

actually the noise is just getting subdued and/or averaged out via the camera's image processor due to the vast amount of information you end up with when having more than 4 times the pixels in the same surface area than the cameras we had 10 years ago

[jsherman999]

...actually the noise is just getting subdued and/or averaged out via the camera's image processor due to the vast amount of information you end up with when having more than 4 times the pixels in the same surface area than the cameras we had 10 years ago

That's pretty much a general-language re-stating of what the author just said in that point, except you left out the technical method used to 'subdue'

[Ikarus]

From this recent dpreview article on the Sony A7 RII (LINK)
"need more bit-depth to properly quantize both the bright and dark signals generated by large pixels.

This is just another way of saying that the larger pixel has the higher dynamic range, so if there is an actual argument in this for smaller pixels being beneficial for reducing noise, it is not immediately obvious. Yes, more bit depth is needed for capturing a higher dynamic range, but more converters are needed for the higher resolution, so this is more a question about efficiency, not so much about noise. It seems to me the real explanation is much simpler - the A7R II benefits from using a more advanced sensor generation. The interesting question to me is whether it would be possible to maintain the high ISO advantage of the A7S' larger pixels over the A7R's first iteration with the latest sensor technology.

[Simen1]

wow, this really exists ?!? sounds amazing, if someone can tweak the Pentax firmware to do the same @_@

Yes, Magic Lantern can do dual ISO, but it only makes sense because of Canon's flat DR-curve between ISO 800 and 100. On Pentax, Nikon and Sony the DR curve is increasing almost linear when reducing ISO from 800 to 100. That means that Canon looses information in the dark areas as ISO is turned down, while Pentax and others don't. I.E. You can do dual ISO in post processing with Pentax, but not with Canon. With the magic lantern approach you loose info in alternating dark and light areas for every other line. You don't loose that info with genuine high DR in all lines on the sensor in Pentax cameras.

Modern sony sensors are close to the ultimate goal of ISO-less behavior. That means you can shoot at base ISO and get the same image quality when pushing up the exposure in post processing as you get from selecting a high ISO in the first place. The advantage of ISO-less behavior is that you can underexpose by many steps to preserve highlight detail and still get as much low light details as you would by selecting a high ISO.

[Chippedoff]

Debatable between DR and Megapixels. There are several examples out there that you can get a clear, sharp image but splotches and bleeding with backlit sensors, the panasonic fz70 is a good example of such flaws. There is always bleeding at the contrast points such as the edges of black and white which end up bleeding into each other and it looks like spilled paint even viewing at a distance. More dynamic range would help but it still won't help the over charge spill over that comes with backlit sensors. The Sony RX100 mark 4 sensor has stepped filters in front of the sensor so it filters out extraneous light. The stepped filter sensor has some old style low light CCD level shadow capture that can be brought out in lightroom with almost no noise increase. Normal backlit sensors have problems with hair or fur where the edge between the hair and the background will often get filled in with blocky jpeg artifact like fog and actually bleed into the hair or background. I held a standard 58mm uv haze filter in front the the fz70 lens and it cleared up about half the fuzzyness outdoors on indoor fluorescent light included.

If the backlit sensor is not pre-filtered it won't matter how good the DR or number of Mpix it has.

[Ratcheteer]

except you left out the technical method used to 'subdue'

no point, camera manufactures are tight lipped on what methods their imaging chips use so anything i say would be a guess and also off topic

anyhow, wish everyone would stop rattling on about sensor noise. keep it simple DR or megapixel

[patarok]

Hybrid Viewfiinder!!! Would be so cool!

Youtu.be

Youtu.be

-- Video from the Fraunhofer Institute about T-OLEDs...

Youtu.be

Youtu.be

-- Another video to a more dedicated OLED..... (years ago...)

Youtu.be

Youtu.be

-- MIT vision for the future of transparent displays...

Youtu.be

Youtu.be

-- another one from a company called "PLANAR" at a fair

(the flickering you see in the vids comes from the video frame rate... One will not perceive it in real life... )