Canon EOS 1DX announced - March 2012 availability

[Ellis Vener]

As I understand it tilting doesnt need a larger imaging circle...

That depends on how much you tilt.  From my years with a 4x5 and 8x10 camera it isn't that hard to tilt either the sensor or lens plane to a point where a corner or two will be out of the imaging circle and some shift is needed to getthe sensor/ film area completely back inside the cone.

[phila]

Thank you for sharing, I did not know. Do these "blockers" have a similar function to the lense hood, or is it simply "putting stuff inside a cramped space where it wont affect IQ on a 24x36mm anyways"? In the first case, it would perhaps be trivial for Canon to re-release their entire EF lense line without "blockers" at minimal delay/cost. In the latter case, that wont be possible.

But it would seem that at least the tilt/shift lenses are exceptions.

-h

They are designed to stop non image forming light from 'bouncing around' inside the lens, which then fall on the sensor degrading the image. Each lens has its own size, design and positioning of 'blocker' according its optical design and they are an integral part of that optical design - so no quick & easy fix. As I said not all lenses have them (although most do). I'm unsure about the T/Ss, it may be possible that they do have them but of a size that does not affect the larger image circle.

[hjulenissen]

They are designed to stop non image forming light from 'bouncing around' inside the lens, which then fall on the sensor degrading the image. Each lens has its own size, design and positioning of 'blocker' according its optical design and they are an integral part of that optical design - so no quick & easy fix. As I said not all lenses have them (although most do). I'm unsure about the T/Ss, it may be possible that they do have them but of a size that does not affect the larger image circle.

This is my source:
http://www.luminous-landscape.com/reviews/cameras/hartblei-cam.shtml

Imagine a camera that can take 35mm format lenses in the front, and medium format digital backs on the back.
...

The above graphic shows the relationship between the size of 135 format and the image circle needed to cover it, and the 645 format and the larger image circle needed to cover that format.
...
what I wasn't prepared for was that the Canon 24mm T/S not only could cover the 645 format but allowed significant movements without visible vignetting.
...
There are likely quite a few other lenses that will work with the HCam. ... whether any particular lens will cover the 645 format is something that you'll have to discover for yourself.

So... perhaps we will see a MF canon back using EF lenses. I still think it would be neat to have a mirror-less "5Dmk3" with a square sensor filling the image circle of the most suited EF lenses to the edges targeted for landscape, portrait and macro, probably also video, not competing with the 1DX (simpler AF, less rugged, no battery grip etc). As Canon does sensors, lenses and everything else themselves (?), this might be one area where it is hard for the competitors to match them.

36x24 represents a 1.77x crop from 60x45, slightly more than the 1.62x crop from 36x24 to 22.2x14.8 if my numbers are right. But within that same 645 image circle, you should be able to fit a 53x53mm square sensor if you have no mirror/ovf, and 24x36 is a whopping 1.8x crop from that :-)

(assuming that "crop" means taking the square root of area1/area2)

Even if that meant that only a few, expensive, specialized EF lenses would currently give full benefit. Other EF lenses should work at least as good as they do today on FF, and Canon might update existing designs/add new specialized ones within the "EF" envelope. If Canon are able to increase sensor size without increasing cost/decreasing other qualities too much, that is.

-h

[alfonsobatalla]

I think stephen is right and he knows a lot about this...
I use Ds MKIII mostly for architectural work with T&S lenses (both Canon and Hartblei). Prints from 150 cm to more than 300 cm.
...and Luminous landscape knows a lot about difraction!
FF sensor @f:11 can only give 18mp (even if you have a highger MP count), a 24 mm (T&S or not) focus from 1 mtr to infinite (hiperfocal at 3m). So, thats all about architecture. Level your camera, focus at 3 mtr, f/11 and there you go, 18 real mp or around 32-36 if you stich pictures with your T&S lens. So, why should I ever need more MP? why should I use MF sensors where depth of field is more restricted?
If canon DX is able to have less noise and better latitude and tonal range, thats the way to go. And I dont mean less noise at very High ISO, but low noise with long exposures, from 30 segs to minutes, where those expensive MF sensors start to get hot and be noisier than point and shot cameras.
Mine is in the waiting list. Its gonna be a long wait.

[Josh-H]

FF sensor @f:11 can only give 18mp (even if you have a highger MP count)

Thats a pretty blanket statement. I'd appreciate if you backed that up with some evidence (as much as for my own education as anything else).

[BernardLanguillier]

Thats a pretty blanket statement. I'd appreciate if you backed that up with some evidence (as much as for my own education as anything else).

Josh,

That's probably about correct. f11 is pretty far from optimal aperture already. f6.3 or 7.1 is the aperture I am trying to use when attempting to extract the highest possible amount of detail from the D3x. f8 starts to be a tiny little less sharp already. That's true with most lenses.

The sad truth is that T/S or DoF stacking are the only 2 ways around trying to get more than 16-18MP worth of resolution from a 35mm sensor when DoF matters. There is no free lunch.

cheers,
Bernard

[Josh-H]

Josh,

That's probably about correct. f11 is pretty far from optimal aperture already. f6.3 or 7.1 is the aperture I am trying to use when attempting to extract the highest possible amount of detail from the D3x. f8 starts to be a tiny little less sharp already. That's true with most lenses.

The sad truth is that T/S or DoF stacking are the only 2 ways around trying to get more than 16-18MP worth of resolution from a 35mm sensor when DoF matters. There is no free lunch.

cheers,
Bernard

Thanks Bernard, I understand the principles of it - and I am generally shooting at F5.6 to F8 as well for exactly the same reasons. However, I admit to not understanding the math behind it - I just reached my own conclusions from shooting with FF High MPX DSLR's with Canons best glass and examining the results.

Assuming for the sake of the argument - We have have a stellar 50mm lens that is critically sharp on a full frame DSLR - shot at F5.6; what is the theoretical maximum resolution that this combination will resolve?

And I ask this as a leading question - If the answer is sub 25 mega pixels; then really... what is the point of cramming more pixels onto the sensor? Are Canon 'right' in making this 1DX an 18 MPX FF sensor based on the resolving power of the lens / sensor combination?

As I say, I am not across the math / engineering - but am curious as to the theoretical limits on the 35mm FF sensor.

[BernardLanguillier]

Thanks Bernard, I understand the principles of it - and I am generally shooting at F5.6 to F8 as well for exactly the same reasons. However, I admit to not understanding the math behind it - I just reached my own conclusions from shooting with FF High MPX DSLR's with Canons best glass and examining the results.

Assuming for the sake of the argument - We have have a stellar 50mm lens that is critically sharp on a full frame DSLR - shot at F5.6; what is the theoretical maximum resolution that this combination will resolve?

And I ask this as a leading question - If the answer is sub 25 mega pixels; then really... what is the point of cramming more pixels onto the sensor? Are Canon 'right' in making this 1DX an 18 MPX FF sensor based on the resolving power of the lens / sensor combination?

Would need to do a bit of math indeed, but if I recall past readings correctly, at f4-f5.6 the theoretical point beyond which more resolution would only produce very limited value was around 40MP.

f5.6 might stop to be optimal beyond 30 MP already, f4 is probably better in the high 30+ megapixels.

You start to run into other limitations of lenses that usually peak between 5.6 and 8... this will justify the design of a new set of lenses optimized for f4...

So all in all, it does probably not make much sense to go beyond 36mp on 35mm. This leaves of course the door open to better pixels, including RGB ones. So it is pretty clear that the winner in 3 to 4 years from now if going to be the company able to improve the usage of a pixel in terms of quality color information. Unless somebody else goes there, the Foveon technology and the related pattents could become worth a lot of dough.

You might of course not have to have 3 colors per pixel though... it could be a good start to do a 2 layers sensor with a smart positioning of RGB sensors on these 2 layers. Jesus... why am I giving away this genius idea for free on a forum?  You have read it here first!

Cheers,
Bernard

[Josh-H]

Would need to do a bit of math indeed, but if I recall past readings correctly, at f4-f5.6 the theoretical point beyond which more resolution would only produce very limited value was around 40MP.

f5.6 might stop to be optimal beyond 30 MP already, f4 is probably better in the high 30+ megapixels.

You start to run into other limitations of lenses that usually peak between 5.6 and 8... this will justify the design of a new set of lenses optimized for f4...

So all in all, it does probably not make much sense to go beyond 36mp on 35mm. This leaves of course the door open to better pixels, including RGB ones. So it is pretty clear that the winner in 3 to 4 years from now if going to be the company able to improve the usage of a pixel in terms of quality color information. Unless somebody else goes there, the Foveon technology and the related pattents could become worth a lot of dough.

You might of course not have to have 3 colors per pixel though... it could be a good start to do a 2 layers sensor with a smart positioning of RGB sensors on these 2 layers. Jesus... why am I giving away this genius idea for free on a forum?  You have read it here first!

Cheers,
Bernard

Thanks Bernard, appreciate the thoughts - interesting times ahead. Im still surprised that neither Canon or Nikon have announced anything above their current high rez offerings in terms of pixel count. I'm quite keen to get my hands on a 1DX and see how it performs in terms of 'pixel quality'. Its not really suited to my style of shooting since I am generally an ISO100, tripod mounted, mirror lock up kind of guy - but my curiosity is getting the better of me. 

[allegretto]

Would need to do a bit of math indeed, but if I recall past readings correctly, at f4-f5.6 the theoretical point beyond which more resolution would only produce very limited value was around 40MP.

f5.6 might stop to be optimal beyond 30 MP already, f4 is probably better in the high 30+ megapixels.

You start to run into other limitations of lenses that usually peak between 5.6 and 8... this will justify the design of a new set of lenses optimized for f4...

So all in all, it does probably not make much sense to go beyond 36mp on 35mm. This leaves of course the door open to better pixels, including RGB ones. So it is pretty clear that the winner in 3 to 4 years from now if going to be the company able to improve the usage of a pixel in terms of quality color information. Unless somebody else goes there, the Foveon technology and the related pattents could become worth a lot of dough.

You might of course not have to have 3 colors per pixel though... it could be a good start to do a 2 layers sensor with a smart positioning of RGB sensors on these 2 layers. Jesus... why am I giving away this genius idea for free on a forum?  You have read it here first!

Cheers,
Bernard

Bernard,

Isn't the real limitation dynamic range?

After all, if the problem with high pixel counts (or perhaps more properly pixel density) is that one gets too much noise. Small pixels have low S/N ratios?

So if pixels could be made with wider dynamic range (thus higher S/N ratios) a manufacturer could mimic the retina and just densely pack pixels of RGB flavor in whatever the optimal pattern and ratios turn out to be.

But perhaps there are other issues too?

[BernardLanguillier]

Isn't the real limitation dynamic range?

After all, if the problem with high pixel counts (or perhaps more properly pixel density) is that one gets too much noise. Small pixels have low S/N ratios?

So if pixels could be made with wider dynamic range (thus higher S/N ratios) a manufacturer could mimic the retina and just densely pack pixels of RGB flavor in whatever the optimal pattern and ratios turn out to be.

Well, a one year old Pentax K5/Nikon D7000 sensor already has as much DR as a P45+ that was described 2 years ago as a DR king, so I don't think it is a problem. A 36mp sensor on FF would have same pixel size as the D7000 sensor.

Regards,
Bernard

[allegretto]

Well, a one year old Pentax K5/Nikon D7000 sensor already has as much DR as a P45+ that was described 2 years ago as a DR king, so I don't think it is a problem. A 36mp sensor on FF would have same pixel size as the D7000 sensor.

Regards,
Bernard

I'm sorry for being obtuse, but isn't the rap on high pixel chips their noise and lack of EV range? This would seem to point to low DR (or at least low S/N ratios) for the smaller pixels. But again, maybe I just don't understand all parameters. Perhaps a better overall question is;

what are current limitations on pixel design? and,

it appears that the Big Two have decided that 24 X 36 chips need no more than 20 Mp plus or minus a couple Mp. They must have decided (separately?) that more than that leads to other compromises that current technology can't surmount

[BernardLanguillier]

it appears that the Big Two have decided that 24 X 36 chips need no more than 20 Mp plus or minus a couple Mp. They must have decided (separately?) that more than that leads to other compromises that current technology can't surmount

Unless you have insider information about the strategy of Canon/Nikon/Sony, I am not sure how you can make this claim.

The Sony A77 at 24mp on a APS sensor has pixels that would result in a 50+ mp if extrapolated to FF. According to our friend Michael Reichmann, the image quality of these Sony is excellent and he is known to be very picky about DR as result of his intense use of MFDB.

I agree wholeheartedly with you that DR is the most important metric of modern sensors, but technology moves forward quickly and 30+MP on a FF sensor is totally doable these days.

Cheers,
Bernard

[alfonsobatalla]

mmm, It seems that I have awaken the difraction monster before going to bed. Well I see some kind people have given so help to my statement
if you trust luminous landscape:
http://forums.dpreview.com/forums/read.asp?forum=1032&message=39627756
and detaliled explanation here
http://www.luminous-landscape.com/tutorials/understanding-series/u-diffraction.shtml
http://www.cambridgeincolour.com/tutorials/diffraction-photography.htm
I have using Ds MKIII for some years now, and 300x100 cm pitures have been hung, exhibited, analized and eventualy sold. for these pano format prints 2 pictures taken with T&S lens were stiched. Nobody ever asked for more resolution.
But i know i´d have a wider latitude, more detail on the shadows, less noise with long exposures and greater tonal range, but not at the expense of loosing my much valuede life preview+life histogram for acurate exposures, or FF DOF for instance. No temptation to move to MF for architecture at all.
So moving from 21 to 18 MP when I actualy have 18 Mp at f:11 is more than OK for me.
On teh other hand if a (for comparision purposes) 16 mp square sensor would only double resolution of a 4 Mp one (yes, 4x4 sensor is twice 2x2 as resolution is taken as a lineal value), what is the real difference betwen 21 and 18?.
Peanuts.
I am counting days down to march.

[allegretto]

Unless you have insider information about the strategy of Canon/Nikon/Sony, I am not sure how you can make this claim.

The Sony A77 at 24mp on a APS sensor has pixels that would result in a 50+ mp if extrapolated to FF. According to our friend Michael Reichmann, the image quality of these Sony is excellent and he is known to be very picky about DR as result of his intense use of MFDB.

I agree wholeheartedly with you that DR is the most important metric of modern sensors, but technology moves forward quickly and 30+MP on a FF sensor is totally doable these days.

Cheers,
Bernard

Well both of their flagships will sport 18-24 Mp chips before long

Maybe I misunderstand but aren't those crowded chips subject to more ISO related noise, thus effectively less DR? But I must be mistaken somewhere?

[BernardLanguillier]

Well both of their flagships will sport 18-24 Mp chips before long

Maybe I misunderstand but aren't those crowded chips subject to more ISO related noise, thus effectively less DR? But I must be mistaken somewhere?

Yep, base ISO DR and high ISO ability are not directly related as some of our chips experts will be able to explain to you better than me.

As far as the Canon and Nikon flagship due to be 18mp cameras, I am not sure where that is coming from.

Their sports camera will probably be 18mp (clear for Canon, unclear for Nikon), but that is already an important increase from the previous model that was 12mp on the Nikon side. This does not mean that another camera with higher resolution will not be released to address markets other than sports.

Cheers,
Bernard

[hjulenissen]

Assuming for the sake of the argument - We have have a stellar 50mm lens that is critically sharp on a full frame DSLR - shot at F5.6; what is the theoretical maximum resolution that this combination will resolve?

And I ask this as a leading question - If the answer is sub 25 mega pixels; then really... what is the point of cramming more pixels onto the sensor? Are Canon 'right' in making this 1DX an 18 MPX FF sensor based on the resolving power of the lens / sensor combination?

Because someone might want to use larger apertures than f/5.6? Because oversampling means that one makes the best possible use of the (expensive) lense, that AA-filtering is easier/not needed, because deconvolution perhaps can recover some of the details?

The combination of sensel density and lense diffraction tends to affect resolution in a smooth fashion, not in a absolute limit, I believe. When someone claim to be "diffraction limited", the point is not that more megapixels will give exactly zero benefit, but that the benefit will be smaller and smaller, finally being too small to be of practical use.

-h

[BJL]

... isn't the rap on high pixel chips their noise and lack of EV range?

No, that is the rap on smaller pixels vs larger pixels only when comparing images of equal pixel count. Like with a smaller sensor of equal pixel count, or in 100% pixel peeping where images from equal sized sensors of different pixel counts are compared by cropping to a smaller portion of the higher resolution image, enlarging it more, and the effectively complaining about the greater visibility of noise and such which is in fact caused by the greater degree of enlargement. That is, editing away all the advantages of the higher pixel count!

AFAIK, comparisons using prints of equal size viewed from the same distance, or other comparisons at "equal apparent image size" do not show any particular trend towards more noise or worse shadow/ highlight handling from sensors of equal size and higher resolution (more, smaller photosites).

Short answer: read about dithering and over sampling.

[allegretto]

No, that is the rap on smaller pixels vs larger pixels only when comparing images of equal pixel count. Like with a smaller sensor of equal pixel count, or in 100% pixel peeping where images from equal sized sensors of different pixel counts are compared by cropping to a smaller portion of the higher resolution image, enlarging it more, and the effectively complaining about the greater visibility of noise and such which is in fact caused by the greater degree of enlargement. That is, editing away all the advantages of the higher pixel count!

AFAIK, comparisons using prints of equal size viewed from the same distance, or other comparisons at "equal apparent image size" do not show any particular trend towards more noise or worse shadow/ highlight handling from sensors of equal size and higher resolution (more, smaller photosites).

Short answer: read about dithering and over sampling.

would love a reference. do you have one handy?

just so I understand... wouldn't a smaller pixel capture fewer photons? and therefore transmit fewer pieces of information?

[allegretto]

Yep, base ISO DR and high ISO ability are not directly related as some of our chips experts will be able to explain to you better than me.

As far as the Canon and Nikon flagship due to be 18mp cameras, I am not sure where that is coming from.

Their sports camera will probably be 18mp (clear for Canon, unclear for Nikon), but that is already an important increase from the previous model that was 12mp on the Nikon side. This does not mean that another camera with higher resolution will not be released to address markets other than sports.

Cheers,
Bernard

I'm just watching from the sidelines. I understand Physics but have few clues about specific chip designs.

D3x = 24 mp

1Dx = 18 Mp

reps say, "... that's all you need for acceptable quality..."

again, not what I say, or even actually know, just what I've read.

Yes, you averred in a previous post that a company could put > 30 Mp on a 24 x 36 chip. But they would be smaller, and have a smaller photon "basket" if current technology was simply scaled.

In a way this takes me back to my initial though which is that a given pixel's ability to transduce photons to electrons is likely currently limited by whatever the limits of current "ideal design" yields. My initial thoughts on this is that chips (pixels) of perhaps a yet unused material may be able to trap more electrons per unit of material and therefore improve resolution (due to recruitment) as well as S/N

But again I'm being theoretical.

Does N or C have a 30+ Mp design in a 24 x 36 chip that actually is "better" from a image and noise perspective? I mean, yes they can do it, but if it brings more issues, why bother?