50 Megapixel Canon in 2015...

(I see John's replied as I was typing, but hopefully this might still contain something useful)

xpfloyd wrote in post #17365431
Exposure and stops I understand - dBs, normalizing, signal noise ratios etc I do not

To steal from Wikipedia: The decibel (dB) is a logarithmic unit used to express the ratio between two values of a physical quantity, often power or intensity."

One confusion is that a 3dB increase is twice the power, but you need 6dB for twice the voltage. IRC in terms of audio, there are claims that 3dB is an actual doubling, but 6dB is perceived as twice the volume.

Normalising in this context I assume to mean taking two different systems and adjusting them to the same scale so a comparison is meaningful. E.g is 16 out of 20 better than 37 out of 50? Scale them both to be n out of 100 and you can compare.

Signal to noise ratio is exactly what it says on the tin; how much signal do I have in relation to noise? Darker ("quieter") pixels are more susceptible to being drowned out by background noise (hence a poor signal to noise ratio). Brighter ("louder") pixels have a signal well above the noise level and thus have good signal to noise.

John Sheehy wrote in post #17365343
Actually, Exmor blackframes are very prone to thin horizontal banding at base ISO. It is exposed by the low amount of random noise.

I'll resist the temptation to make a smart*ss comment about banding on Exmor sensors

BTW when I said difference I meant (practically) "visible" as opposed to "measurable". I don't know if that would change the answer.

Thanks John and sploo for the explanations, I'll read them a few more times and hope it sinks in

Reservoir Dog wrote in post #17365419
Have no worries, anyways we do not take 5 stops under-exposed pictures to push them back later on the computer to see what we got

No, but you might take a photo that's correctly exposed for the highlights, but which has so much dynamic range that the shadows need to be pushed 5 stops to reveal detail.

An indoor shot (no flash possible, for various reasons) with windows on a sunny day comes to mind. As do many landscapes.

Shadowblade wrote in post #17365934
No, but you might take a photo that's correctly exposed for the highlights, but which has so much dynamic range that the shadows need to be pushed 5 stops to reveal detail.

An indoor shot (no flash possible, for various reasons) with windows on a sunny day comes to mind. As do many landscapes.

Knowing that i can recover in RAW 3 stops on each side of the histogram without banding i will expose in the "middle" to be able to recover on both side (yea i should say 3 stops the big maximum after it's ugly)

Do not not make me wrong, i understand that you enjoy the "Tec" part (i do too), i understand the purpose of your discussion, and i'm learning those spec with this interesting discussion, and i do agree being able to recover 5 stops on just one side of the histogram without banding like the Sony sensor, would be a real improvement in some situation you named or in my case > underwater where sometime i can got headache to expose properly and more headache to recover some shadows.
but for everyday situation i'm actually very happy

3 stops head room on highlights? I've seen maybe 1.5 with raw files from any of my digital cameras but not 3

smythie wrote in post #17366215
3 stops head room on highlights? I've seen maybe 1.5 with raw files from any of my digital cameras but not 3

English is not my mother language, i don't understand what you mean by "head room".
But on the histogram (LR or DPP) it's 3 stops on the left and 3 stops on the right side too before there is nothing else.

When I say head room, I mean that if highlights in a portion of the image look blown out, there is often a stop of so of exposure that can be pulled from that area or the image as a whole while keeping detail in those highlights. I haven't been able to retrieve more than that from my highlights. I don't pay much attention to the position on the histogram though so can't provide a reference point for you.

smythie wrote in post #17366435
When I say head room, I mean that if highlights in a portion of the image look blown out, there is often a stop of so of exposure that can be pulled from that area or the image as a whole while keeping detail in those highlights. I haven't been able to retrieve more than that from my highlights. I don't pay much attention to the position on the histogram though so can't provide a reference point for you.

Ok i got it, thank you.
I do look at the histogram, not so often i should do , though, but it tell me if i can recover something or not, beyond a certain point

I'd like a medium format sensor please.

Gel wrote in post #17367095
I'd like a medium format sensor please.

Well, you're in luck: Phase One and Hasselblad have your mf sensor needs covered.

Seriously, if Canon made an mf camera, wouldn't it require a whole new lens system? Or can they make EF lenses work with a larger sensor?

And if you need all new lenses, why not go with already-existing well-established mf ecosystems?

Reservoir Dog wrote in post #17366198
Knowing that i can recover in RAW 3 stops on each side of the histogram without banding i will expose in the "middle" to be able to recover on both side (yea i should say 3 stops the big maximum after it's ugly)

I think that might illustrate a misunderstanding with regard to the term "recover". My understanding of "recovering" shadows is the process of raising the brightness of the darkest areas of the image (the darkest stops). If the level of noise in those dark pixels is sufficiently low, I'll be able to obtain usable (visually acceptable) detail. Obviously there will come a point at which parts of the scene are too dark to have recorded any useful information (or are now below the noise floor, so contain no usable detail).

At the highlights end, anything that's brighter than the saturation level of a pixel in the sensor is clipped - no detail to recover. The camera's histogram may show areas as being clipped because they often base the info on the generated jpeg; whereas the raw data will likely contain data in pixels that hasn't yet clipped - hence the impression that you're "recovering" blown highlights.

Ideally, a digital camera would expose a scene so that the brightest parts (in which we still want detail) are just below the saturation/clipping level, and all other (darker) detail is now hanging down from that upper level. If the sensor has very low noise, and the digital path has sufficient bits, then we'll get plenty of stops of usable detail (well into the dark shadows) that can be manipulated in post.

If you expose a scene "in the middle" then more of your highlights will be pushed into clipping; though of course in some scenarios this may be what you wanted (shadow details favoured over highlights).

Note also that banding should only affect the left side of the histogram; where the signal is low, and is affected by pattern noise resulting from the camera's electronics. As you head towards the right side the signal levels will be way above the pattern noise, so it shouldn't be a problem.

Reservoir Dog wrote in post #17366198
Knowing that i can recover in RAW 3 stops on each side of the histogram without banding i will expose in the "middle" to be able to recover on both side (yea i should say 3 stops the big maximum after it's ugly)

There aren't 3 extra stops to "recover" in the highlights in normal circumstances. With Daylight WB, a typical conversion uses a level 2.5 stops above metered middle gray for "white", and in a "normal contrast" conversion, and everything above that clips away. Depending on the camera model and ISO, there is generally 2.5 to 4 stops above middle gray in the green channel, which is 0 to 1.5 stops of extra highlights. The natural color balance of our sensors is such that in a grayscale scene in white light, the RAW red channel is about 1 stop less sensitive than green, and blue is about 1/2 stop less sensitive, so this actually gives another stop of highlights for the red channel, and a half for the blue. So, the color channels generally clip separately for all intents and purposes. In any area of the image where one or two channels are clipped, it is impossible to get full-color detail; only grayscale is possible, though some converters allegedly try to guess the color, based on surrounding areas with no clipping. There is never more than 1.5 stops of full-color information above normally-clipped white, except in some camera that obtain certain ranges of ISOs by "under-exposing" a lower ISO, or when using HTP (highlight tone priority) on a Canon, which under-exposes ISO 100 by a stop to get ISO 200 with HTP.

There is nothing equivalent to highlight clipping on the shadow side of a RAW's tonal levels. The RAW records signals down to no signal at all, which is an infinite number of stops below white or gray. The only problem is, there is so much noise that it is hard to recognize details, especially fine ones, and noise that has patterns in it is far worse at hiding real signal. If your task was not aesthetic (like making a "nice" photo"), and you simply wanted to do something like identify a bold shape that nearly filled the frame, like a large letter, or a large arrow and you wanted to see which way it pointed, you could literally go many stops lower into the shadows at ISO 1600 and identify the shape, if you know how to process it properly, even with a P&S camera's RAW.

Gel wrote in post #17367095
I'd like a medium format sensor please.

Are you ready to finance the new set of lenses you will need? The only thing you might carry over of any value is a speedlight, or wireless flash transmitter, or remote releases. Not much cheaper than buying into an existing MF platform, as the biggest expenses are bodies and lenses.

There's no such thing as 'head room' in the highlights of a digital RAW file. If you're in the highlights of the RAW file, you're already very close to clipping. If you're not, then your highlights aren't in the true highlights of the RAW file, but merely appear to be since default JPEG conversion tends to clip the brightest 1-2 stops of the RAW file to white. With regards to dynamic range, the only thing that matters is how many stops below 'blown out' you can still resolve detail above the noise floor, not how many stops above an arbitrary midpoint you can go.

I just wish you could save a custom curve to the camera for in-camera JPEG conversion and to use as the setting for display on the camera's LCD. That way, you could set a curve that showed the full DR of the image, instead of the clipped default curve, and could better utilise the full DR of the camera instead of just guessing how many stops above the default curve's 'white' you actually are and, therefore, how close you are to blowing out the RAW.

David Arbogast wrote in post #17367110
Well, you're in luck: Phase One and Hasselblad have your mf sensor needs covered.

Seriously, if Canon made an mf camera, wouldn't it require a whole new lens system? Or can they make EF lenses work with a larger sensor?

And if you need all new lenses, why not go with already-existing well-established mf ecosystems?

This medium format talk is likely off topic because Canon is so unlikely to actually come out with a medium format body, but if they did, Canon TS-e 17, 24, 45 and 90mm would all cover the senor, and that alone would be a great set, even if there was little room for movement. Two somewhat longer lenses would do most of what anyone would likely want to do with the camera. Well...., except AF.

But the other poster is right- I would have to finance them.