End of the APS-H sized CMOS?

[Bob Howland]

A top Nikon executive in a interview for DPR stated that they had made a big mistake in sacrificing resolution for high ISO performance in the D3S. He stated that future models would not make this same error.

So, even Nikon feels that there is more advantage to higher resolution than to the ultimate high ISO performance.

If we're thinking of the same quote, that's not exactly what he said. First, he didn't call it an error. Second, he implied that Nikon felt it could begin increasing resolution at the expense of further improving high ISO image quality. The expectation among Nikon users seems to be that the D4 will be as good as the D3s at high ISO but with more resolution. I know a couple wedding photographers who each use a pair of D700s (plus a D300 backup, but that's another story). They have become addicted to what the D700 can do at ISO6400 and say they would like to buy the D3s, but their business won't support it.

 

[Haydn1971]

Ivar, I'm not gonna get into a tit-for-tat cut and paste exercise, it's clear you disagree with my point of view. You simply ain't gonna get the same framerates with full frame as a crop frame, the files are smaller with a crop, which is why there will always be a market for a high end crop body.

If the Nikon range is so good, maybe you need to sell your Canon kit and swap to the dark side ;-)

 

[x-vision]

Why do you think Canon should stop APS-H sensors? What is the interest for Canon?

Because the 1.3x sensor is a lesser sensor than FF. This means that:

• if Canon charges more for the 1DIV than Nikon for the D3s, Canon will be overcharging for a lesser camera
  
• if Canon charges less (as is the case today), the long term perception will be that Nikon is the premium player in the market and Canon is the discount/second grade player

As you can see, Canon has no choice but to abandon the 1.3x format - unless they want to be the perceived as the discount/second grade player in the market for pro cameras.

 

[Haydn1971]

Because the 1.3x sensor is a lesser sensor than FF.

Nonsense ! Each has it's own advantages, it's bold sweeping statements like this that confuse new buyers and like saying a BMW M5 is a lesser car than a Rolls Royce Phantom, the M5 is better at certain things, the Rolls better at others.

 

[Loswr]

Because the 1.3x sensor is a lesser sensor than FF.

Correct - in terms of the area dimension of the sensor, APS-H is smaller than FF. So what? "Full frame' is the lesser sensor when compared to MF - maybe we should all harass Phase One and Hasselblad for a 10 fps camera (or, at least, to measure frame rate in frames per second instead of seconds per frame...)?!?

APS-H meets the needs of some photographers, and not others. It's always best to use the right tool for the job, and Canon supports that with a variety of sensor formats.

Regarding the comparisons to Nikon, I don't think those are really relevant at the top end of the line. The market fraction that chooses one of the highest-end bodies (1D/1Ds, D3x/D3s) as their first camera must be infinitesimal. Those who upgrade/replace into the flagship series have already bought into a system, and if they are changing systems, a cost difference in the bodies is a fraction of the total cost (new lenses, etc.). As it stands now, the difference in price between the 1DIV and D3s is insignificant (4%), and the difference between the 1DsIII and D3x is still not that meaningful (14%).

 

[Ivar]

I haven't argued what you stated - if it is possible to process only x megapixels at 10fps, then you would rather have all "the pixels" in a denser form though in a smaller sensor area.

The question is, if there is more processing power (file, data, megapixels etc as you wish), whether it makes sense to make a FF 27MP camera or make a 1.3x 27MP camera (mechanically the ability is already there, it is only about the data)

I believe currently, that with today's technology, the pixel density doesn't matter for high ISO, but the size of the light capture area does - more pixels might not contribute to the higher ISO, but the overall quality is never less than bigger pixels at the same sensor area, for both at the same size print. This is the basis for the choice - would you like to have a better ISO or more magnification, being under the data processing constraint?

Ivar, I'm not gonna get into a tit-for-tat cut and paste exercise, it's clear you disagree with my point of view. You simply ain't gonna get the same framerates with full frame as a crop frame, the files are smaller with a crop, which is why there will always be a market for a high end crop body.

If the Nikon range is so good, maybe you need to sell your Canon kit and swap to the dark side ;-)

 

[Haydn1971]

I haven't argued what you stated - if it is possible to process only x megapixels at 10fps, then you would rather have all "the pixels" in a denser form though in a smaller sensor area.

The question is, if there is more processing power, whether it makes sense to make a FF 27MP camera or make a 1.3x 27MP camera.

The 1.3 crop isn't the same mp as the full frame, the pixel size is about 10% smaller, 5.7 vs 6.4, there isn't a current 1.3 crop that gives the same mp as a current full frame. Because there are less pixels, there is less processing time required for each frame. The 1.3 crop currently has about 25% less data to deal with, thus is faster, the pixel size is similar size, so you aren't actually losing that much except frame size.

If 1.3 crop was the same mp as a full frame, the pixel size would be about that (perhaps less) of a APS-C, which would give you just a single benefit of being able to take slightly wider angle shots than a APS-C with the same lens, but not much else, which would of course be too small a reason to spend £10's Millions in developing a specific 1.3 crop.

 

[Ivar]

The 1.3 crop isn't the same mp as the full frame, the pixel size is about 10% smaller, 5.7 vs 6.4, there isn't a current 1.3 crop that gives the same mp as a current full frame. Because there are less pixels, there is less processing time required for each frame. The 1.3 crop currently has about 25% less data to deal with, thus is faster, the pixel size is similar size, so you aren't actually losing that much except frame size.

Absolutely true. With disagreement only in the very last sentence after the last comma. This is where the potential is not yet used by Canon.

If 1.3 crop was the same mp as a full frame, the pixel size would be about that (perhaps less) of a APS-C, which would give you just a single benefit of being able to take slightly wider angle shots than a APS-C with the same lens, but not much else, which would of course be too small a reason to spend £10's Millions in developing a specific 1.3 crop.

Let's try to go to the basics - do you agree that, provided using the same (and I mean the same) technology , the bigger light capture area has better signal, thus theoretically better IQ?

 

[Haydn1971]

Let's not go back to basics - currently, there aren't similar mp full frame and 1.3 crop sensors, so your point is meaningless, which is why there are currently two different types of sensor and why it's unlikely that the 1.3 crop will go away.

 

[Rocky]

If 1.3 crop was the same mp as a full frame, the pixel size would be about that (perhaps less) of a APS-C, which would give you just a single benefit of being able to take slightly wider angle shots than a APS-C with the same lens, but not much else, which would of course be too small a reason to spend £10's Millions in developing a specific 1.3 crop.

Let's try to go to the basics - do you agree that, provided using the same (and I mean the same) technology , the bigger light capture area has better signal, thus theoretically better IQ?

"Same Technology", I am assuming you mean same pixel density. The picture quality (noise) per pixel is identical regardless of the acyual sensor size. The FF will have more pixel than the APS-H and hence better picture quality. Hydyn1971 is half right about the comparision of APS-C ans APS-H with the same pixel density. besides a widen angle with the same lens, The APS-H will also have a higher resolution and hence better opicture quality.

 

[Bob Howland]

If 1.3 crop was the same mp as a full frame, the pixel size would be about that (perhaps less) of a APS-C, which would give you just a single benefit of being able to take slightly wider angle shots than a APS-C with the same lens, but not much else, which would of course be too small a reason to spend £10's Millions in developing a specific 1.3 crop.

Let's try to go to the basics - do you agree that, provided using the same (and I mean the same) technology , the bigger light capture area has better signal, thus theoretically better IQ?

"Same Technology", I am assuming you mean same pixel density. The picture quality (noise) per pixel is identical regardless of the acyual sensor size. The FF will have more pixel than the APS-H and hence better picture quality. Hydyn1971 is half right about the comparision of APS-C ans APS-H with the same pixel density. besides a widen angle with the same lens, The APS-H will also have a higher resolution and hence better opicture quality.

It's entirely possible for two sensors, designed and built years apart, to have the same pixel pitch/size but use different "technologies". Moreover, it is possible for the newer sensor to have smaller pixels but the same number of them as the older and, because of technology improvements, provide a better image quality.

 

[Rocky]

[It's entirely possible for two sensors, designed and built years apart, to have the same pixel pitch/size but use different "technologies". Moreover, it is possible for the newer sensor to have smaller pixels but the same number of them as the older and, because of technology improvements, provide a better image quality.

We are talking about "same technology" here. That will imply the same semiconductor processing for the sensors. Who with the right mine will put a smaller pixel size with the same sensor with the same MP count even with newer technoloy. Canon is doing the absolute opposite. It make the pixel as big as posible (for better noise performanance), use gapless microlens, at least on the 7D officially, may even be with the Rebels, for better light gethering and hence better noise performance. I have nor heard of any body decrease the pixel size in the same sensor with the same pixel count. May be you can quote me an example.

 

[Bob Howland]

[It's entirely possible for two sensors, designed and built years apart, to have the same pixel pitch/size but use different "technologies". Moreover, it is possible for the newer sensor to have smaller pixels but the same number of them as the older and, because of technology improvements, provide a better image quality.

We are talking about "same technology" here. That will imply the same semiconductor processing for the sensors. Who with the right mine will put a smaller pixel size with the same sensor with the same MP count even with newer technoloy. Canon is doing the absolute opposite. It make the pixel as big as posible (for better noise performanance), use gapless microlens, at least on the 7D officially, may even be with the Rebels, for better light gethering and hence better noise performance. I have nor heard of any body decrease the pixel size in the same sensor with the same pixel count. May be you can quote me an example.

You can't be serious. Canon is making the pixels as large as possible? That'll surprise a lot of people. Please compute the pixel sizes of the Nikon D700, Canon 5DMk2, Nikon D300S and Canon 7D. The percentage of the pixel which actively gathers light, the design of the microlenses, sensor fabrication methods, the number of sensor readout channels, the post processing after the analog signals leave the sensor and the quality of the A-to-D converter all could be classified as "technology". Are you and Hayden1971 deliberately trying to confuse matters and create non sequiturs. REMOVED. Lets be CIVIL

 

[Rocky]

[It's entirely possible for two sensors, designed and built years apart, to have the same pixel pitch/size but use different "technologies". Moreover, it is possible for the newer sensor to have smaller pixels but the same number of them as the older and, because of technology improvements, provide a better image quality.

We are talking about "same technology" here. That will imply the same semiconductor processing for the sensors. Who with the right mine will put a smaller pixel size with the same sensor with the same MP count even with newer technoloy. Canon is doing the absolute opposite. It make the pixel as big as posible (for better noise performanance), use gapless microlens, at least on the 7D officially, may even be with the Rebels, for better light gethering and hence better noise performance. I have nor heard of any body decrease the pixel size in the same sensor with the same pixel count. May be you can quote me an example.

You can't be serious. Canon is making the pixels as large as possible? That'll surprise a lot of people. Please compute the pixel sizes of the Nikon D700, Canon 5DMk2, Nikon D300S and Canon 7D. The percentage of the pixel which actively gathers light, the design of the microlenses, sensor fabrication methods, the number of sensor readout channels, the post processing after the analog signals leave the sensor and the quality of the A-to-D converter all could be classified as "technology". Are you and Hayden1971 deliberately trying to confuse matters and create non sequiturs. REMOVED. Lets be CIVIL

Yes I am serious. I work in the semiconductor industry on the techical area. I know what technology means. We are talking about pixel size in the sensors only. Nothing has been memtioned about anything outside the sensor. Please read the the technical papers. 7d has the highest pixel density density. It will have the smallest pixel pitch,. However, Canon is trying to make the indvidual pixel as large as possible and use Gapless microlens to artifficial to increase the pixel size. FYI, number of channel for the sensor read out is not technology, it is "implementation".
You just keep on throwing in irrelevent points to confuse the issue. Until you can show me an example of camera maker actually decrease the pixel size for a given sensor with the same pixel count, you are just arguing for the sake of argueing. Look, who is not civilized.

 

[Ivar]

Let's try to go to the basics - do you agree that, provided using the same (and I mean the same) technology , the bigger light capture area has better signal, thus theoretically better IQ?

"Same Technology", I am assuming you mean same pixel density. The picture quality (noise) per pixel is identical regardless of the acyual sensor size. The FF will have more pixel than the APS-H and hence better picture quality. Hydyn1971 is half right about the comparision of APS-C ans APS-H with the same pixel density. besides a widen angle with the same lens, The APS-H will also have a higher resolution and hence better opicture quality.

There are quite many variables currently, we need to use ceteris paribus principle ie fix some variables to understand each other.

Let's define couple of situations:
1) Let's suppose the same density PLUS the same technology used as a different technology with the same density can give quite a lot of variations in image quality:
1a) a camera with a cropped sensor
1b) a FF camera

2) Let the density vary meaning both, the crop camera and the FF will have N megapixels, again using the same technology.:
2a) a camera with a cropped sensor
2b) a FF camera

Now some thoughts:
3. For the first two cameras, 1a and 1b, yes, no problems to understand the benefits of a larger sensor IQ wise as it just records more data on a bigger surface (some characteristics may however degrade, for example fps when being constrained by the data processing capability ie MPs per second).
4. For the second case, let's suppose that for a given print at least 300ppi is available, I claim that ISO-wise the camera 2b is much better on paper.

The fear for APS-H advocates seems to be:
5. by being under data processing constraint, the crop camera, having the same number of megapixels, would end up with higher density thus better "reach"
6. FF is more expensive

In my opinion, at certain level of density in combination with high ISO (meaning fine detail is being lost), the benefit is questionable (look at the compact small sensor cameras, the density for SLR-s is going towards that). This is why I think it is time to change to FF in order to provide competitive IQ.

I also, let's be dramatic, hate lenses acting as being handicapped. It is a big deal for example 24-105 not acting like one. After all, the world is not composed of tele lenses only.

 

[Rocky]

Ivar, There is no doubt that FF is ALWAYS better than APS_C or APS-H. I am using a APS-C (40D) with a 17-40mm as my main lens. It will serve me 90% of the time. The other 10% (special situation) will be the 35-135( for longer reach), 18-55mm Kit lens with IS ( for hand held low light) and very occationally 70_300mm DO( It is heavy). I am not a fan of telephoto lens. The reason I use the APS-C is the cost, size and weight is less than the FF. I like to have the APS-H if the size and price will come down. APS-H will make my 17mm becomes 22mm and get rid of the funny behavior of the lens at the corners the same time. I think Canon should make a 7D size APS-H body with the price to match.

 

[AJ]

What if:

Canon were to start making FF sensors with two pixel densities: one very dense (perhaps 32 or 36 mpix) for those 3-4 fps 5D and 1Ds series cameras, and a lower density (maybe 24 or 18) mpix sensor for a 1D series camera.

The lower pixel density FF sensor would give unprecedented high-iso performance and allow huge fps. It'd be great for sports, journalism, things like that. It'd be great for video too!

Question then is of cost. Right now APS-H sensors are only put into 1D cameras. Would it be cheaper for Canon to produce two formats only? Would a low-density FF sensor cost significantly more than a APS-H sensor with similar pixel count?

 

[hlphoto]

What if Canon didn't focus on very dense FF sensors nor APS-H...?

I think Canon should make a 7D size APS-H body with the price to match.

But I'm afraid you're about the only one who really wants this. Well, there might be a few people.. but Canon doesn't gain a lot by serving themselves with another cheap competitor for a more expensive camera like they did with the 5D Mk2 / 1Ds Mk3 earlier...

"Same Technology", I am assuming you mean same pixel density. The picture quality (noise) per pixel is identical regardless of the acyual sensor size. The FF will have more pixel than the APS-H and hence better picture quality. Hydyn1971 is half right about the comparision of APS-C ans APS-H with the same pixel density. besides a widen angle with the same lens, The APS-H will also have a higher resolution and hence better opicture quality.

Only partly true. Everybody seems to be forgetting that the bigger a sensor grows it also grows very different thermal specifications. I'm 99% sure that if you grow the 7D sensor to 56x36mm and still have it run at ISO 6400 (native), it would drive you mad with noise and kill your batteries in a snap(shot).

Same goes for FF: The 5D classic is very close to the 5D Mk2 in terms of noise (in print, not per pixel), but has some trouble with amp noise, especially if you push it (high ISO, slow shutter, in hot danceclubs like I used to do).

That said... I think APS-H has to come to an end too. I regularly shoot the 1D Mk3, 1D Mk4, both the first and second 5D, and sometimes a 1Ds Mk3. Despite it is somewhat an older camera I really like the 1D Mk3, because it feels really solid, is extremely fast and serves me with more than decent ISO 2500 shots which can be denoised quite nicely. The 5D Mk2 makes cleaner images, but is not worth it's 21mp of data at ISO 2500 concerning detail. The 10.1mp files out of the 1D are really detailed.

Just yesterday I took the 1D Mk4 and the 5D Mk2 with me for a shoot... edited the pictures last night. You would expect the 1D Mk4 to beat the 5D Mk2 in terms of noise, right?

WRONG. The 1D Mk4 might have a little less noise, but the structure of this noise and the way it damages colours is horrible. Also: It is very hard to remove this noise! Lightroom, Photoshop, Capture One and DPP all can't really handle the 1D sensor output like they do the 5D2.
I found myself dropping loads of ISO 5000 shots out of the 1D4 while using several ISO 5000 shots out of the 5D and even a ISO 12800 shot (although it needed a lot of work, especially the blue channel gets baaad when you push it this hard).

Out of all Canons cameras, if you'd make me mix & match all parts into a new body, I would say make a 1D out of:
- give it the 1Ds Mk3 AF (which I've found to be more reliable than the 1D Mk4 AF when things get really dark - Canon be ashamed!);
- update the 1Ds Mk2 sensor to todays standards - ISO 12800, gapless microlenses, decent DR/colour rendition etc. (I even like the 5D Mk1 high ISO "grain" better than that of the 1Ds Mk2, but 12.8mp isn't enough for the near future, even though it is for my standards);
- higher ISO than 12800 is a possibility.. but tell me, does anybody EVER use more then ISO 12800? I did once and it took ages to edit those pictures. The only reason to do so was for a stop more DOF, and if I was there again I wouldn't even try. Cut the crap Canon, forget the high ISO race;
- where is that rumoured DIGIC 5? Give us two. I bet handling 12 frames of 16.7mp per second wouldn't be a single problem;
- ISO 25 would be nice, but at least 50 without losing DR;
- Would like to see an AF point controller which is actually reachable without squeezing my enormous hand between my head and cambody when in portrait mode - this is NOT fast at all;
- give Nikon a run for their money by pushing 12fps - we all know both the mirror assembly and shutter should be capable, Nikon already managed 11 with the D3. Is it useful? No. Is it fun? Sure

I would price my camera somewhere between the 5000 and 6000 dollar mark, and make it possible to buy something like "CPS Plutonium" (in stead of "CPS Platina"): say 750 dollars for 1 day exchange-service and free quarterly calibration through the first two years. That is pro service like pro service should be. Enables hobbyists to get a massive camera for less money than pros doe and enables pros to concentrate on their work.

Also... make a simple version of this camera with 3 fps, less fancy AF, the slower mirror/shutter assembly of the 5D Mk2, simpler (only centered?) light metering, maybe even plastic body like the 60D etc. and call it 5D Mk3. Price it around 3500 dollars (yes, it will sell). Make sure both are capable of very high spec video (maybe full resolution?) - the 5D has a very big set of fans in both high-end prosumer photographers and videographers.

I bet a camera like "my 1D" would be a winner for both studio and sports. No 1Ds needed. If you're such a hotshot and really need your megapixels... buy medium format!

 

[Mt Spokane Photography]

What if:

Canon were to start making FF sensors with two pixel densities: one very dense (perhaps 32 or 36 mpix) for those 3-4 fps 5D and 1Ds series cameras, and a lower density (maybe 24 or 18) mpix sensor for a 1D series camera.

The lower pixel density FF sensor would give unprecedented high-iso performance and allow huge fps. It'd be great for sports, journalism, things like that. It'd be great for video too!

Question then is of cost. Right now APS-H sensors are only put into 1D cameras. Would it be cheaper for Canon to produce two formats only? Would a low-density FF sensor cost significantly more than a APS-H sensor with similar pixel count?

Cost of a sensor is relative to the size, pixel count is not important. Canon has a good white paper explaining this.

http://www.scribd.com/doc/38542576/Canon-Cmos-Wp

Production costs for a full-frame sensor can exceed twenty times the costs for an APS-C sensor. Only 20 full-frame sensors will fit on an 8-inch (200 mm) silicon wafer, and yield is comparatively low because the sensor's large area makes it very vulnerable to contaminants—20 evenly distributed defects could theoretically ruin an entire wafer. Additionally, the full-frame sensor requires three separate exposures during the photolithography stage, tripling the number of masks and exposure processes.[

http://en.wikipedia.org/wiki/Full-frame_digital_SLR

Pixel densities on FF cameras is pretty low compared to 1.6 crop, and extremely low compared to point and shoot cameras, so pixel density is relative.

Density of the 7D would yield a 46mp FF, so 36mp would be like a 14mp crop camera.

And if a FF had the density of the 10mp G12 and S95, it would have about 200mp!

So FF has a long ways to go to catch up density wise.

Of course, large pixel counts result in large files which are hard to manipulate in editing software.

 

[Loswr]

I have nor heard of any body decrease the pixel size in the same sensor with the same pixel count. May be you can quote me an example.

Until you can show me an example of camera maker actually decrease the pixel size for a given sensor with the same pixel count, you are just arguing for the sake of argueing.

Note that you and Bob are not talking about the same thing. Sorry, but where did Bob mention "same sensor" or "same sensor size"? Let's see:

It's entirely possible for two sensors, designed and built years apart, to have the same pixel pitch/size but use different "technologies".

Bob is not saying smaller pixels in the same sensor - he's saying you may find the same pixel sizes in different sensors designed years apart (e.g. the 5DII from 2008 and the 20D from 2004 have the same pixel pitch).

I don't think any manufacturer would release a new sensor where the only/main change was a reduction in the pixel size, keeping sensor size and pixel count the same. Obviously, the usual trend is to use smaller pixels in the same sensor size, (e.g. 10 MP APS-C to 18 MP APS-C), and there Canon has done a good job of keeping the overall IQ consistent despite decreasing pixel sizes. Canon has also bucked the usual trend, going from a 12 MP PowerShot G9 to a 14 MP PowerShot G10, but then down to a 10 MP G11/G12, all with the same 1/1.7" sensor.

Moreover, it is possible for the newer sensor to have smaller pixels but the same number of them as the older and, because of technology improvements, provide a better image quality.

Bob is also saying that a new sensor, one with the same pixel count but smaller pixels (but again, not the same size sensor) might have better IQ than the older sensor with larger pixels. If you'd like a specific example (as close as I can find, at any rate), compare the 1Ds (original) with the 1D MkIII - similar MP (11 and 10.1 MP, respectively), but the 1DIIII has ~33% smaller pixels (based on area, pitch is 8.8 µm vs. 7.2 µm), and 5 years of improved technology behind it's smaller pixels. DxOMark shows that the 1DIII bests the 1Ds across the board. For a related example, you could compare the 1DsII with the 1DIV, where the MP count is quite similar (16.6 vs. 16.1 MP), the pixels are ~50% smaller (based on area, pitch is 7.2 µm vs. 5.7 µm), the sensor is smaller, but the IQ is equivalent. In both cases, 5 years of technological advancements compensated for, and in one case improved upon, the handicap of both smaller pixels and a smaller sensor format.