EOS-1D X Mark II Dynamic Range [CR2]

[tpatana]

The C300 mk.ii uses a sensor with similar QE/SN as the 6D. The difference would be the DPAF. Those extra pixels can be used to expose the shade areas for example; expose half the file at +4ev, and the non DPAF pixels expose at 0 EV; then the RAW file is cooked to spread this exposure latitude over one file (i.e Average out normal pixels and the DPAF ones shot at +4Ev. Very much like the ML Dual_iso hack. Except that this will be done at a hardware layer, meaning no loss of resolution like in Dual_iso.

The implementation used on the C300 mk II has the output from each pixel simultaneously fed into two amplifiers - one for the shadows/mid tones, and the other for mid tones/highlights. This way each segment of the picture sees the entire captured dynamic range, not split up into alternate pixels, and much more importantly, not split into alternate phases.

Let me understand this better, The amplifier reading the shadows would need to amplify more since the pixel is not going to collect more photons beyond the shutter closing up. This would denote an operation similar to high iso (i.e. Signal amplification), so would that introduce noise?... or perhaps noisier than the mid to highlight capture which requires less amplification?

For a given number of photons captured you want to read out at the *highest* possible ISO, not the lowest. This will give you the lowest noise. The reason low ISOs have less noise is that more light is captured, not that the electronics are less noisy.

Thanks, I had it backwards. I thought hi-ISO meant high amplification, thus higher noise....

That's how I thought it too. Lee Jay, can you give more details to explain how that works? In most electronics, increasing amplification also increases noise. What's the difference here?

 

[Aglet]

But I don't have the fast-focusing long lenses for BiF shooting, it's not my area of interest to put that kind of loot into a big lens I don't have a use for.

You really should consider getting a supertele lens. Unlike smaller lenses where the lens cap merely sits on the front of the lens, supertele lens cap/covers fit completely over the front and extend down the sides like a glove, which ensures complete blockage of any light entering the lens. That could have a profoundly beneficial impact your lens cap photography.

at the risk of giving away my proprietary lens-cap testing methodology, that full sleeve front cap won't help a bit.
But, how many stops of DR are required for BiF pictures? ;D

 

[exquisitor]

The C300 mk.ii uses a sensor with similar QE/SN as the 6D. The difference would be the DPAF. Those extra pixels can be used to expose the shade areas for example; expose half the file at +4ev, and the non DPAF pixels expose at 0 EV; then the RAW file is cooked to spread this exposure latitude over one file (i.e Average out normal pixels and the DPAF ones shot at +4Ev. Very much like the ML Dual_iso hack. Except that this will be done at a hardware layer, meaning no loss of resolution like in Dual_iso.

The implementation used on the C300 mk II has the output from each pixel simultaneously fed into two amplifiers - one for the shadows/mid tones, and the other for mid tones/highlights. This way each segment of the picture sees the entire captured dynamic range, not split up into alternate pixels, and much more importantly, not split into alternate phases.

Let me understand this better, The amplifier reading the shadows would need to amplify more since the pixel is not going to collect more photons beyond the shutter closing up. This would denote an operation similar to high iso (i.e. Signal amplification), so would that introduce noise?... or perhaps noisier than the mid to highlight capture which requires less amplification?

For a given number of photons captured you want to read out at the *highest* possible ISO, not the lowest. This will give you the lowest noise. The reason low ISOs have less noise is that more light is captured, not that the electronics are less noisy.

Thanks, I had it backwards. I thought hi-ISO meant high amplification, thus higher noise....

That's how I thought it too. Lee Jay, can you give more details to explain how that works? In most electronics, increasing amplification also increases noise. What's the difference here?

The thing with the read noise is a bit more complex. There are three sources of the total read noise: read noise (RN) from the sensor itself, amplification noise (AN) and analog-to-digital converter (ADC) noise. At the low ISO most cameras are ADC noise limited, with the higher ISO ADC noise decreases and after roughly 1600-3200 most cameras become RN limited.
The reason for high ADC noise is (as I understand it) is the high level of the maximal signal at low ISO (corresponding to full well capacity). ADC converters have certain signal-to-noise characteristics, expressed like percentage of the full well capacity signal fed to the ADC. The full well capacity is ISO dependent and is maximal at the lowest ISO.
At low ISO there are a lot of electrons in highlight areas, however in the shadow areas you've got much less electrons. Because all data from the sensor processed in the same way, the signal from the shadow has more noise form ADC. For example if we have ADC with 1 %SNR characteristics, we will get 10 electrons noise at 1000 electrons signal at ISO 100 in highlight area (SNR =100) and ... same 10 electrons noise at 20 electrons signal (SNR = 2). This is why Canon cameras have high shadow noise. Sony's sensors have much less ADC noise.
So now the trick with dual ISO readout is to make two parallel ADC channels and process all data simultaneously at different ISO. At low ISO the highlights will get processed the best way as described above. At high ISO the shadows will be processed with lower full well capacity and respectively with lower maximal signal. So as in the example above, if we have maximal signal of 62.5 electrons at ISO 1600, the noise in the shadows will become 1/16 of that at ISO 100: 0.625 electrons corresponding to SNR of 32.

 

[Lee Jay]

The C300 mk.ii uses a sensor with similar QE/SN as the 6D. The difference would be the DPAF. Those extra pixels can be used to expose the shade areas for example; expose half the file at +4ev, and the non DPAF pixels expose at 0 EV; then the RAW file is cooked to spread this exposure latitude over one file (i.e Average out normal pixels and the DPAF ones shot at +4Ev. Very much like the ML Dual_iso hack. Except that this will be done at a hardware layer, meaning no loss of resolution like in Dual_iso.

The implementation used on the C300 mk II has the output from each pixel simultaneously fed into two amplifiers - one for the shadows/mid tones, and the other for mid tones/highlights. This way each segment of the picture sees the entire captured dynamic range, not split up into alternate pixels, and much more importantly, not split into alternate phases.

Let me understand this better, The amplifier reading the shadows would need to amplify more since the pixel is not going to collect more photons beyond the shutter closing up. This would denote an operation similar to high iso (i.e. Signal amplification), so would that introduce noise?... or perhaps noisier than the mid to highlight capture which requires less amplification?

For a given number of photons captured you want to read out at the *highest* possible ISO, not the lowest. This will give you the lowest noise. The reason low ISOs have less noise is that more light is captured, not that the electronics are less noisy.

Thanks, I had it backwards. I thought hi-ISO meant high amplification, thus higher noise....

That's how I thought it too. Lee Jay, can you give more details to explain how that works? In most electronics, increasing amplification also increases noise. What's the difference here?

Two separate but related things happen when you change ISO. First, you change the sensor analog amplification (usually, this is sometimes also done purely in digital). Second, you change the way the in-camera meter works with higher ISOs meaning shorter or darker exposures. The second one costs you light and that creates a lower signal to noise ratio.

However, what's the point of the additional amplification? The point is to *reduce* noise created in the pipeline between the pixel and the output of the analog-to-digital converter. An unfortunate side effect of that is that it means that big signals will saturate the ADC creating clipped highlights.

So, for a given exposure you want the highest gain and thus the lowest noise you can get that clips only as many highlights as you are willing to tolerate being clipped.

As always, more (brighter) exposure will get you more light captured and thus lower noise and that's why setting a lower ISO generally gets you a lower-noise photo, but that's not what I was talking about. I was talking about *for a given exposure* and for that you want the highest analog gain you can get.

What Canon is doing here is trying to get the lowest noise possible (highest ISO) but not accepting all those clipped highlights and therefore sampling everything at a low gain (low ISO) at the same time, and blending the two to get a shot without clipped highlights but with low shadow noise as well.

 

[tpatana]

Ok, I think I got what you meant. So in short, ETTR gives less noise for same picture conditions than if you exposure left, hence higher ISO = less noise.

 

[pedro]

The C300 mk.ii uses a sensor with similar QE/SN as the 6D. The difference would be the DPAF. Those extra pixels can be used to expose the shade areas for example; expose half the file at +4ev, and the non DPAF pixels expose at 0 EV; then the RAW file is cooked to spread this exposure latitude over one file (i.e Average out normal pixels and the DPAF ones shot at +4Ev. Very much like the ML Dual_iso hack. Except that this will be done at a hardware layer, meaning no loss of resolution like in Dual_iso.

The implementation used on the C300 mk II has the output from each pixel simultaneously fed into two amplifiers - one for the shadows/mid tones, and the other for mid tones/highlights. This way each segment of the picture sees the entire captured dynamic range, not split up into alternate pixels, and much more importantly, not split into alternate phases.

Let me understand this better, The amplifier reading the shadows would need to amplify more since the pixel is not going to collect more photons beyond the shutter closing up. This would denote an operation similar to high iso (i.e. Signal amplification), so would that introduce noise?... or perhaps noisier than the mid to highlight capture which requires less amplification?

For a given number of photons captured you want to read out at the *highest* possible ISO, not the lowest. This will give you the lowest noise. The reason low ISOs have less noise is that more light is captured, not that the electronics are less noisy.

Thanks, I had it backwards. I thought hi-ISO meant high amplification, thus higher noise....

That's how I thought it too. Lee Jay, can you give more details to explain how that works? In most electronics, increasing amplification also increases noise. What's the difference here?

Two separate but related things happen when you change ISO. First, you change the sensor analog amplification (usually, this is sometimes also done purely in digital). Second, you change the way the in-camera meter works with higher ISOs meaning shorter or darker exposures. The second one costs you light and that creates a lower signal to noise ratio.

However, what's the point of the additional amplification? The point is to *reduce* noise created in the pipeline between the pixel and the output of the analog-to-digital converter. An unfortunate side effect of that is that it means that big signals will saturate the ADC creating clipped highlights.

So, for a given exposure you want the highest gain and thus the lowest noise you can get that clips only as many highlights as you are willing to tolerate being clipped.

As always, more (brighter) exposure will get you more light captured and thus lower noise and that's why setting a lower ISO generally gets you a lower-noise photo, but that's not what I was talking about. I was talking about *for a given exposure* and for that you want the highest analog gain you can get.

What Canon is doing here is trying to get the lowest noise possible (highest ISO) but not accepting all those clipped highlights and therefore sampling everything at a low gain (low ISO) at the same time, and blending the two to get a shot without clipped highlights but with low shadow noise as well.

Very intresting. Even as a non-tech guy I kinda get the idea. Will this trickle down into a 5DIV or 5DX or whatever may come out next? I don't mind if it is 2016 or 17, my 5D3 still rocks...

 

[Woody]

DxO is scrambling to come up with a different rating system in order to slam Canon sensors after reading this news.

They don't have to. Their sensor scores are so cryptic that they can assign any number they like without any justification.

 

[K-amps]

Two separate but related things happen when you change ISO. First, you change the sensor analog amplification (usually, this is sometimes also done purely in digital). Second, you change the way the in-camera meter works with higher ISOs meaning shorter or darker exposures. The second one costs you light and that creates a lower signal to noise ratio.

However, what's the point of the additional amplification? The point is to *reduce* noise created in the pipeline between the pixel and the output of the analog-to-digital converter. An unfortunate side effect of that is that it means that big signals will saturate the ADC creating clipped highlights.

So, for a given exposure you want the highest gain and thus the lowest noise you can get that clips only as many highlights as you are willing to tolerate being clipped.

As always, more (brighter) exposure will get you more light captured and thus lower noise and that's why setting a lower ISO generally gets you a lower-noise photo, but that's not what I was talking about. I was talking about *for a given exposure* and for that you want the highest analog gain you can get.

What Canon is doing here is trying to get the lowest noise possible (highest ISO) but not accepting all those clipped highlights and therefore sampling everything at a low gain (low ISO) at the same time, and blending the two to get a shot without clipped highlights but with low shadow noise as well.

Thanks... sort of reminds me of the old analog amplification days with in-line limiters, where you'd have a zener diode in line with the signal, which when saturated would increase negative feedback thus reducing overall output so that the next stage (e.g preamp input) would not be saturated. In the end, the effect is not to increase DR, rather manage is by decreasing DR to a given architecture (e.g. 14bit or 16bit encoding).

 

[K-amps]

The thing with the read noise is a bit more complex. There are three sources of the total read noise: read noise (RN) from the sensor itself, amplification noise (AN) and analog-to-digital converter (ADC) noise. At the low ISO most cameras are ADC noise limited, with the higher ISO ADC noise decreases and after roughly 1600-3200 most cameras become RN limited.
The reason for high ADC noise is (as I understand it) is the high level of the maximal signal at low ISO (corresponding to full well capacity). ADC converters have certain signal-to-noise characteristics, expressed like percentage of the full well capacity signal fed to the ADC. The full well capacity is ISO dependent and is maximal at the lowest ISO.
At low ISO there are a lot of electrons in highlight areas, however in the shadow areas you've got much less electrons. Because all data from the sensor processed in the same way, the signal from the shadow has more noise form ADC. For example if we have ADC with 1 %SNR characteristics, we will get 10 electrons noise at 1000 electrons signal at ISO 100 in highlight area (SNR =100) and ... same 10 electrons noise at 20 electrons signal (SNR = 2). This is why Canon cameras have high shadow noise. Sony's sensors have much less ADC noise.
So now the trick with dual ISO readout is to make two parallel ADC channels and process all data simultaneously at different ISO. At low ISO the highlights will get processed the best way as described above. At high ISO the shadows will be processed with lower full well capacity and respectively with lower maximal signal. So as in the example above, if we have maximal signal of 62.5 electrons at ISO 1600, the noise in the shadows will become 1/16 of that at ISO 100: 0.625 electrons corresponding to SNR of 32.

Thank you Sir. Can't say I am ready for a quiz, but I get the idea.

 

[exquisitor]

Very intresting. Even as a non-tech guy I kinda get the idea. Will this trickle down into a 5DIV or 5DX or whatever may come out next? I don't mind if it is 2016 or 17, my 5D3 still rocks...

There was a rumor that this tech will come to 5D IV. Now it's used in C300 II.
However it's not clear, if the dual ISO will be used in 1DX II. It could be a new ADC design that has lower noise. IMO dual ISO is still sounds more plausible. What ever the reason is, Canon is avoiding to use on-chip-in-row ADC like that found in Sony's sensors.

 

[exquisitor]

The thing with the read noise is a bit more complex. There are three sources of the total read noise: read noise (RN) from the sensor itself, amplification noise (AN) and analog-to-digital converter (ADC) noise. At the low ISO most cameras are ADC noise limited, with the higher ISO ADC noise decreases and after roughly 1600-3200 most cameras become RN limited.
The reason for high ADC noise is (as I understand it) is the high level of the maximal signal at low ISO (corresponding to full well capacity). ADC converters have certain signal-to-noise characteristics, expressed like percentage of the full well capacity signal fed to the ADC. The full well capacity is ISO dependent and is maximal at the lowest ISO.
At low ISO there are a lot of electrons in highlight areas, however in the shadow areas you've got much less electrons. Because all data from the sensor processed in the same way, the signal from the shadow has more noise form ADC. For example if we have ADC with 1 %SNR characteristics, we will get 10 electrons noise at 1000 electrons signal at ISO 100 in highlight area (SNR =100) and ... same 10 electrons noise at 20 electrons signal (SNR = 2). This is why Canon cameras have high shadow noise. Sony's sensors have much less ADC noise.
So now the trick with dual ISO readout is to make two parallel ADC channels and process all data simultaneously at different ISO. At low ISO the highlights will get processed the best way as described above. At high ISO the shadows will be processed with lower full well capacity and respectively with lower maximal signal. So as in the example above, if we have maximal signal of 62.5 electrons at ISO 1600, the noise in the shadows will become 1/16 of that at ISO 100: 0.625 electrons corresponding to SNR of 32.

Thank you Sir. Can't say I am ready for a quiz, but I get the idea.

You're welcome! If you want get more details on this topic and generally how the sensor works, I would recommend this article: http://www.clarkvision.com/articles/digital.sensor.performance.summary/index.html .

 

[pedro]

Thank you, exquisitor. In reference to improved high ISO IQ what will the improvement be like, on correctly exposed images using on-chip-in-row-ADC in comparison to the discussed dual ISO concept? Or, what wil be the overall improvement of either of those new concepts over current ADC tech? 1/2, 1, or two stops combined with new Digics and reasonable MP count? I would be absolutely happy with 12800ish ISO 25.600 ;-)

 

[jrista]

Thank you, exquisitor. In reference to improved high ISO IQ what will the improvement be like, on correctly exposed images using on-chip-in-row-ADC in comparison to the discussed dual ISO concept? Or, what wil be the overall improvement of either of those new concepts over current ADC tech? 1/2, 1, or two stops combined with new Digics and reasonable MP count? I would be absolutely happy with 12800ish ISO 25.600 ;-)

High ISO performance is going to be more dependent upon sensor Q.E. than ADC performance. Because the signal is initially amplified in each pixel, it is stronger coming off the sensor. The ADC will still add noise, but it's a relatively consistent amount of noise, so relative to a strongly amplified signal, it's small. There is also little value in using dual-ISO techniques at high ISO, as you aren't able to make full use of the dynamic range of the sensor in the first place.


As Q.E. on current Canon sensors is already up in the 60% range, it's unlikely we will see a 1-stop improvement in high ISO noise unless larger pixels are used (or something else that increases well capacity). It is primarily at low ISO where a reduction in read noise will increase dynamic range. A dual-ISO approach can certainly improve things, however as your effectively blending two different exposures with different noise characteristics, this can often create artifacts (check MagicLantern images). The best approach is to actually use better technology to prevent noise in the first place. Isolating high-frequency components away from ADC units, using more ADC units so they can operate at lower frequency, reducing trace distance from pixel to ADC, converting to digital at the earliest convenience and using error-corrected data transfer, etc.

 

[exquisitor]

Thank you, exquisitor. In reference to improved high ISO IQ what will the improvement be like, on correctly exposed images using on-chip-in-row-ADC in comparison to the discussed dual ISO concept? Or, what wil be the overall improvement of either of those new concepts over current ADC tech? 1/2, 1, or two stops combined with new Digics and reasonable MP count? I would be absolutely happy with 12800ish ISO 25.600 ;-)

You are welcome.
High ISO capability is not limited by ADC noise anymore and thus ADC type is not relevant in this case. The noise at high ISO comes from actual read noise from the sensor and from amplification noise. However these two guys are already very good at this point and there is not much room for improvement. So to achieve cleaner high ISO there are two ways:
1). improve quantum efficiency of the sensor to use more light arrived to the pixel;
2). make larger pixel to gather more light.
1Dx has about 50 % quantum efficiency, so there is a room for improvement. The resolution would be IMO in the 20-24 MP range. Overall I think you could expect at least half a stop better high ISO, optimistically even 1 stop. But that is just speculation...

 

[pedro]

Thanks a lot to both of you jrista and exquisitor for the helpful explanation! If a 5DIVc came out at 18 MP I'd go for it, as long as one can shoot stills as well, without severe limitations. As for the pixel size there would be less noise due to more light gathering if I am correct. As I mostly do astro and lowlight, this one would do just fine for me. Or will a regular 5DIV benefit from a 1/2 stop IQ improvement as well due to the new tech? I hope so. Wish they'd leave it in the low 20 MP, and increase to 24 MP max. 22 MP and an improved sensor would be even better of course! 8)

 

[jrista]

Thanks a lot to both of you jrista and exquisitor for the helpful explanation! If a 5DIVc came out at 18 MP I'd go for it, as long as one can shoot stills as well, without severe limitations. As for the pixel size there would be less noise due to more light gathering if I am correct. As I mostly do astro and lowlight, this one would do just fine for me. Or will a regular 5DIV benefit from a 1/2 stop IQ improvement as well due to the new tech? I hope so. Wish they'd leave it in the low 20 MP, and increase to 24 MP max. 22 MP and an improved sensor would be even better of course! 8)

There are different kinds of noise. There is noise in the signal itself, and noise added by the electronics. Bigger pixels mean more light. Signal grows faster than noise in the signal, so with bigger pixels, you get a higher SNR. Technically speaking, however, noise is also growing, not shrinking, with bigger pixels...it's just that with the higher SNR, our perception of it changes.


If you have small pixels that can gather 5000e- at half well (midtone gray), then the noise in that signal would be SQRT(5000), or 70.71e-. Now, lets say you double the pixel size, in which case it would have four times the area, and thus be capable of gathering four times as much light in the same exposure (time the shutter is open at a given aperture). You now have 20,000e- at half well. The noise in that signal is SQRT(20000), or 141.42e-. The noise is MORE, not less...however the signal is much much more. The ratio between signal and noise, or S/N, which is really S/SQRT(S), increased. You have an SNR of 5000/SQRT(5000) for the smaller pixels, which is 70.71:1, and an SNR of 20000/SQRT(20000) for the larger pixels, which is 141.42:1.


There is also the sources of electronic noise added to the signal. Those are on top of the noise in the signal itself. At ISO 100, let's say the 5D IV has half the noise of it's predecessor. That would be about 16e- RN. Let's say it has similar dark current to the 7D II, in which case (outside of astro, at least) it's meaningless. Our S/N then becomes more complex: S/SQRT(S+RN^2). With our smaller pixels, our SNR is 5000/SQRT(5000+16^2), or 5000/72.5, 68.97:1. With the bigger pixels, our SNR is 20000/SQRT(20000+16^2), or 20000/142.32, which is 140.52:1. Read noise is compounded with the noise in our signal, increasing the overall noise in our images.


With a bright signal, like we get in the midtones, the increase caused by read noise is negligible. However at low signal levels, such as you would have with your astrophotography, the amount of read noise becomes significantly more important. You might have an object signal (at a true dark site, 21.5mag/sq") of 200e- for a five minute exposure with smaller pixels. That impacts our SNRs more. Smaller pixels would have an SNR of 9.36:1, while bigger pixels would have an SNR of 24.61:1. The bigger pixels do have a higher SNR, but it's still low compared to the midtone signal.


---


TL;DR


There is another factor to consider with astro. Image scale. With normal terrestrial photography, you compose your subject relative to your frame, and that is pretty much that. Composition in astrophotography is a bit different, and there is sampling to consider. For a given scope and camera, your field of view is your field of view...your not going to be changing it, as everything is effectively at infinity. Sampling ratio then becomes a significant factor (for DSO imaging, at least...if your doing ultra wide field milky way imaging, this doesn't really apply). How many pixels are being used to represent each star? If stars only cover about one pixel each, then all your stars will be square. Even if you manage to get a 2x2 matrix of pixels covering each star...your stars are still going to be mostly square. Sampling becomes very important to accurately resolving details in astrophotography. You want about 3.5x3.5 pixels sampling each star, at the very least. Picking a camera then becomes an exercise in combining the pixel size of your sensor and the focal length of your lens or scope, to get a more ideal sampling.


Small pixels will sample better at shorter focal lengths. At 300mm, you could well need ~1.5 micron pixels to be well sampled. At 600mm you will need about 3 micron micron pixels. At 1000mm or so, 5 micron pixels are better. At 2000mm, 9 micron pixels are better. A Sony A7s makes for an ideal long focal length camera, as it's cheap (compared to CCD cameras with 9 micron pixels, which cost $8000-$20,000), and has nice, but 7.4 micron pixels. A 7D II would be great for an 800-1000mm scope or lens. An APS-C camera with 28mp would be pretty nice for a 600mm lens or scope. It's actually pretty tough to find sensors with 1.5 micron pixels...so image scale tends to suffer when you get wider.


Pixel size and noise are a different best in astro. You expose multiple sub frames, and technically speaking, more is always better. A lot of astro imagers get 2-3 hours of exposure, and leave it at that. I myself usually get about 10-12 hours. Great imagers are usually working with 20, 30, 60 hours of total exposure time. The amount of noise in a single sub becomes less meaningful the more you integrate...and since you can effectively integrate an infinite amount of data, the amount of noise in astro images ultimately boils down to the individual imager's tolerance for exposing the same target for many nights, and spending the necessary compute cycles to integrate ever larger volumes of data.

 

[K-amps]

Thanks Jon. Your posts are a mouthful, but very helpful. You are generous with your research.

 

[pedro]

Thanks again, jrista!

 

[sanj]

Jrista.
My wish: To one day understand what you wrote.

 

[Diko]

I've never seen such an argumentative, self absorbed, smug, infantile and downright pathetic group as that which regularly posts here. Nothing is more important than being right, starting a fight, arguing over nothing or scoring points.

Good day MR. bollo, that is what discussion groups are all about:
Entertainment, Education & Clash of Egos.

You obviously dislike all that "noise". So feel free to never ever allow thyself wasting any precious time around here.
It is your (birth)right.

...dual pixel AF in the 70d, for example), right?

Nope. 70D otherwise had little to offer. So they needed urgently to come up with something so not to stop the cash flow. ;-) Actually if you have been around. Since the 5D mark II the only ground braking novelty was exactly DAF. Otherwise it was a boring cycle of moderately (compared to the competition) increasing the specs. I don't count the new flash 600 series, because I see nothing genuinely new there either.

I said WAS, because now these rumors of new CMOS with competitive DR and the new E-TTL III are quite promising. As it seems finally Canon has heard its users and came to reality where there exist competition.

Yes you are right, but it is a rumor site. We like to discuss possibilities. One thing leads to another and along the way we learn photography techniques .
....
Besides this is the internet. It needs to be different than a social club. Here faceless people can and should say their minds. That is the fun bit...

That was beautiful... May I quote you?

I want to be able to push shadows so many stops that I can shoot with lens cap on and still recover the picture in LR.

Some traditional Canon users would consider shooting at base ISO in low light with the lens cap still on as bad technique - I consider it one if the many basic situations the sensor should be able to overcome

Absolutely. 95% of the time I begin shooting with the lens cap on. Which also makes always my models smile. A good time for a photo ;-)