Do Sensors Make the Camera?

[Don Haines]

Why is a cooperation like Canon not making a new sensor fabrication?

A new CMOS production line worth at least twenty million dollars. Without enough motivation, a old school cooperation is very unlikely to spend this kind of money on things like that. Therefore, the company decided to work the engineers' buds off to push the limit for the old 500nm sensor because it is cheaper.

The best reason that I can think of as to why they are not buying a new CMOS production line is because with the downturn in P/S camera sales, production capacity is opening up on their existing production line for small sensors, which is a finer line than that used for FF and APS-C (Non DPAF) sensors.

The second point to make is that the complexity of circuitry required to make a DPAF 70D sensor is beyond that which can be fabricated on the 500nM line... therefore, it must be fabricated on something else and that means the existing fabrication line for P/S sensors.... This means that they are now in the process of switching over to a finer line and when that is done, they can shut down the 500nM line and save money..... because it costs more to keep 2 lines running than 1 line.

 

[jrista]

Why is a cooperation like Canon not making a new sensor fabrication?

A new CMOS production line worth at least twenty million dollars. Without enough motivation, a old school cooperation is very unlikely to spend this kind of money on things like that. Therefore, the company decided to work the engineers' buds off to push the limit for the old 500nm sensor because it is cheaper.

The best reason that I can think of as to why they are not buying a new CMOS production line is because with the downturn in P/S camera sales, production capacity is opening up on their existing production line for small sensors, which is a finer line than that used for FF and APS-C (Non DPAF) sensors.

The second point to make is that the complexity of circuitry required to make a DPAF 70D sensor is beyond that which can be fabricated on the 500nM line... therefore, it must be fabricated on something else and that means the existing fabrication line for P/S sensors.... This means that they are now in the process of switching over to a finer line and when that is done, they can shut down the 500nM line and save money..... because it costs more to keep 2 lines running than 1 line.

I think this is probably the most reasonable line of reasoning as to why Canon hasn't moved to a finer fabrication process. It's very logical and simple. Let's just hope they wind down P&D production quickly, so their next DSLR cameras can start putting a finer lithography to use.

 

[Jonasyphoto]

Why is a cooperation like Canon not making a new sensor fabrication?

A new CMOS production line worth at least twenty million dollars. Without enough motivation, a old school cooperation is very unlikely to spend this kind of money on things like that. Therefore, the company decided to work the engineers' buds off to push the limit for the old 500nm sensor because it is cheaper.

The best reason that I can think of as to why they are not buying a new CMOS production line is because with the downturn in P/S camera sales, production capacity is opening up on their existing production line for small sensors, which is a finer line than that used for FF and APS-C (Non DPAF) sensors.

The second point to make is that the complexity of circuitry required to make a DPAF 70D sensor is beyond that which can be fabricated on the 500nM line... therefore, it must be fabricated on something else and that means the existing fabrication line for P/S sensors.... This means that they are now in the process of switching over to a finer line and when that is done, they can shut down the 500nM line and save money..... because it costs more to keep 2 lines running than 1 line.

60D and 70D shares the same iq performance according to DXO mark, and that means the split pixels it does not improve the file quality. Also according to a post on Chiphell about dynamic range, D800e with 180nm Fab sensors are obviously outperforming D4 w' 250nm fab sensor. (I had doubt on it's results due to comparisons such as 5Dmk3 VS D800 and D800 vs. IQ180, after using all of these cameras, I am trusting DXOmark now)

Let's hope Canon is switching or buying a Sony sensor because "even a user knows how to expose correctly" need wider dynamic range and better files from my camera.

Let's knock on wood folks.

 

[3kramd5]

and that means the split pixels it does not improve the file quality.

That's not what he was getting at. He's saying (I believe) they already have a finer process, and citing DPAF as proof, even though they currently are using it for something other than IQ.

 

[Don Haines]

and that means the split pixels it does not improve the file quality.

That's not what he was getting at. He's saying (I believe) they already have a finer process, and citing DPAF as proof, even though they currently are using it for something other than IQ.

Yes!
With the available space on the die, I believe that there is not enough space to build the circuitry required for DPAF for the 70D using the 500nM process...

 

[jrista]

and that means the split pixels it does not improve the file quality.

That's not what he was getting at. He's saying (I believe) they already have a finer process, and citing DPAF as proof, even though they currently are using it for something other than IQ.

Well, to be fair, he is still speculating about the 70D being made with a 180nm process. The 70D has only a few extra gates to handle the ability to read the two separate photodiodes in each pixel separately, as well as to bin them for a full pixel read. There isn't even a 10% increase in pixel FWC...if they had moved to 180nm, that means the border of wiring and transistors around each pixel would shrink by 320nm each side. That would lead to an increase in photodiode area of nearly 43%. Even assuming the independent read and binning logic takes up more space, I would still have expected more than a 9% increase in FWC if the 70D had moved to a 180nm process.

 

[Don Haines]

and that means the split pixels it does not improve the file quality.

That's not what he was getting at. He's saying (I believe) they already have a finer process, and citing DPAF as proof, even though they currently are using it for something other than IQ.

Well, to be fair, he is still speculating about the 70D being made with a 180nm process. The 70D has only a few extra gates to handle the ability to read the two separate photodiodes in each pixel separately, as well as to bin them for a full pixel read. There isn't even a 10% increase in pixel FWC...if they had moved to 180nm, that means the border of wiring and transistors around each pixel would shrink by 320nm each side. That would lead to an increase in photodiode area of nearly 43%. Even assuming the independent read and binning logic takes up more space, I would still have expected more than a 9% increase in FWC if the 70D had moved to a 180nm process.

If they had done DPAF on the 500nM process, I would have expected the FWC to have dropped... yet, as you point out, it has slightly gone up...

of course, all this is speculation..... I have no inside information and am just guessing at a possible scenario....

 

[jrista]

and that means the split pixels it does not improve the file quality.

That's not what he was getting at. He's saying (I believe) they already have a finer process, and citing DPAF as proof, even though they currently are using it for something other than IQ.

Well, to be fair, he is still speculating about the 70D being made with a 180nm process. The 70D has only a few extra gates to handle the ability to read the two separate photodiodes in each pixel separately, as well as to bin them for a full pixel read. There isn't even a 10% increase in pixel FWC...if they had moved to 180nm, that means the border of wiring and transistors around each pixel would shrink by 320nm each side. That would lead to an increase in photodiode area of nearly 43%. Even assuming the independent read and binning logic takes up more space, I would still have expected more than a 9% increase in FWC if the 70D had moved to a 180nm process.

If they had done DPAF on the 500nM process, I would have expected the FWC to have dropped... yet, as you point out, it has slightly gone up...

of course, all this is speculation..... I have no inside information and am just guessing at a possible scenario....

I think it depends on whether they could fit all teh transistors in that 500nm border around the photodiodes or not. If there was room in that space not dedicated to photodiode area, then it's possible.

I wonder if they moved to some intermediary process first...320nm, 250nm?

 

[bosshog7_2000]

I agree that Canon is behind in sensor tech, but like many others already stated it is the camera SYSTEM that is most important. Look at the amazing A7R as an example...awesome sensor but limited lenses and camera functionality.

The other thing to keep in mind is that a $1900 Canon 6D has better image quality than anything out there 7 yrs ago....that said, it would be nice if Canon could re-take the position of superiority in sensor design.

 

[SwampYankee]

Sensors do not make the camera but all other things being close to equal it is a deciding factor. If I had nothing invested in any line it would be hard to find a good reason to buy a Canon 5DIII over a Nikon 810. Every thing else is close, auto focus, build, lens line, accessories, price, historical reliability, 3rd party support . Bit of an edge to Canon on service and support, big edge to Nikon on sensor.

 

[Famateur]

I agree that Canon is behind in sensor tech, but like many others already stated it is the camera SYSTEM that is most important.

If you said, "Canon is behind in sensor dynamic range and low ISO noise," I think I'd agree. With a general term like "sensor tech", well, that leaves it open to which tech you're referring to. Dual-Pixel Auto-Focus is sensor tech, and, as far as I know, Sony/Nikon sensor tech is nowhere close in this regard.

I fully agree with your second, point. While I'd love more dynamic range and better ISO performance, for most things that I shoot, I'd rather "get the shot" and have to do some tweaking in post than have extra dynamic range and editing latitude for a missed shot. Obviously, that wouldn't apply much to landscape shooters...

Of course, the whole dynamic range debate all depends on each photographer's needs, so it's kinda pointless to argue. I doubt Canon decision-makers argue around the board room table about whether dynamic range is important or not (maybe they do). My guess is that they plunk down a fair amount of market research dollars (yen?) to settle that matter and move forward. Not that they conclude dynamic range is unimportant -- just that other things (like DPAF) take priority in the development release pipeline.

I'm in the system-over-sensor* camp -- not because I'm "right", but because that's what works best for me.




*Let me qualify this by specifying current sensors. I'm not suggesting that a really crappy sensor in a good system is a good choice. I also don't consider Canon's current sensors to be crappy. :-X

 

[Jonasyphoto]

Okey, let me express myself a bit more clear. Canon is definitely able to make high-res sensors, it also able to make amazing prototypes in lab.

The file quality improvement in lower ISO from 5Dmk2 to 5Dmk3 is negligible, and I really can not find a reason for it. For this reason I have been thinking about jump ship to Nikon or go MF for a while. A camera packed with 2012 funtion and 2007 IQ(1dsmk3) is really disappointing.

I am now almost always use Capture One to convert CR2s just to get the file a bit more retouching friendly.

 

[Maiaibing]

Re: Do Sensors sell the Camera?

...since the 5DII, I haven't seen a Canon camera with a sensor that was significantly better enough for me to want to buy it or recommend it to anyone.

100% spot on!

Now lets have Canon bring on a worth while 5DIV with a sensor on par (I'll be happy with that) with what others are offering...

 

[scyrene]

Okey, let me express myself a bit more clear. Canon is definitely able to make high-res sensors, it also able to make amazing prototypes in lab.

The file quality improvement in lower ISO from 5Dmk2 to 5Dmk3 is negligible, and I really can not find a reason for it. For this reason I have been thinking about jump ship to Nikon or go MF for a while. A camera packed with 2012 funtion and 2007 IQ(1dsmk3) is really disappointing.

I am now almost always use Capture One to convert CR2s just to get the file a bit more retouching friendly.

From what I've read in seemingly dozens of similar debates on here, Canon focused on other things when designing the 5D3 - autofocus and high ISO in particular - because that's what people wanted. I wasn't following things back then, but apparently that's what people were clamouring for - the DR/ultra high resolution/low ISO quality stuff has only come along since then.

Are the shots you've shared meant to show Canon's limitations? They look fine, I don't get it.

 

[dtaylor]

The thing your missing is that detail buried in "shadow" isn't the problem. It's detail buried in READ NOISE that's the problem. "Shadows" extend for hundreds to thousands of levels...read noise usually only intrudes a dozen or so levels into the deepest of shadows. It's those very deep shadows that mark the difference between a Canon sensor and an Exmor.

The image you edited is more along the lines of this:

Your Nikon example is also more along those lines. Your sunflower shot with all the color noise is pitch black before being pushed. The bedroom scene, night scene, and sunflower shot without all the color noise are not pitch black. You can see some shadow detail in all of them.

Honest questions...not trying to pick a fight after the last thread...what RAW converter are you using on the sunflowers? What NR settings did you set in the converter itself (not in post)? Are you willing to provide the "pitch black" RAW file for others to try their hand at?

 

[jrista]

The thing your missing is that detail buried in "shadow" isn't the problem. It's detail buried in READ NOISE that's the problem. "Shadows" extend for hundreds to thousands of levels...read noise usually only intrudes a dozen or so levels into the deepest of shadows. It's those very deep shadows that mark the difference between a Canon sensor and an Exmor.

The image you edited is more along the lines of this:

Your Nikon example is also more along those lines. Your sunflower shot with all the color noise is pitch black before being pushed. The bedroom scene, night scene, and sunflower shot without all the color noise are not pitch black. You can see some shadow detail in all of them.

Honest questions...not trying to pick a fight after the last thread...what RAW converter are you using on the sunflowers? What NR settings did you set in the converter itself (not in post)? Are you willing to provide the "pitch black" RAW file for others to try their hand at?

The Nikon image has pitch black shadows. It's the back sides of the bed, the back side of the desk, and the back side of the curtain. It was over a +5 stop lift, according to the guy who made the images.

As for my file, have at it:

http://1drv.ms/1vmTXXq

 

[RLPhoto]

No, but they are part of it.

 

[MichaelHodges]

I have, but the majority of my photography is birds and wildlife. That's all high ISO, high frame rate, heavy AF based work, and the 5D III is a much, much better camera for that than the 6D. The frame rate of the 5D III is actually a bit slow...I could really use the 8fps of the 7D combined with all the 5D III features and larger frame.

I agree the 5D III has the superior AF for wildlife, but the 6D gets the job done, even for very fast animals.

The 5D III can be used for landscapes. It's just a lot more work to clean up each image in post than I had expected it to be. Based on the raving reviews and everyone who said the 5D III noise was much better, I expected the 5D III low ISO noise to be a lot better than it is. Even a +3 stop shadow recovery is actually asking a bit much of the 5D III...you still get vertical banding when you do that, and the 5D III banding does not seem to clean up as well as the 7D did. I put off landscapes for a while, as I did not like dealing with the shadows on the 7D at ISO 100 (and, because the frame size in the 7D is just not large enough for the kinds of landscapes I like to do.) Now that I have the 5D III, it's still not capable of doing quite what I want to do...so I'm considering buying a D810 and a 14-24mm f/2.8.

If the 5D III is not cleaning up as well as your 7D, there is something very wrong with your 5D III, or there's a setting off in your software. I use Lightroom for processing, and with the most recent version, the difference between Canon FF and 7D detail recovery is massive.

Although I will say the shadow recovery on my 6D is quite good. Maybe not Nikon/Sony level, but certainly superior to APS-C.

I could wait for Canon to do something with the 5D IV...but, I don't really think they will. I think the 5D IV will arrive with the same old kind of read noise, they will still be losing two stops of dynamic range...and I don't think it's worth waiting any longer.

Yeah, I think it's clear Canon is going to sit tight with their conservative designs. More and more they are becoming like Toyota, competent and comfortable, but living off their peak of 2001-2005.

 

[Keith_Reeder]

Re: Do Sensors sell the Camera?

...since the 5DII, I haven't seen a Canon camera with a sensor that was significantly better enough for me to want to buy it or recommend it to anyone.

Learn how to convert them properly, then.

70D, 160 ISO, can go from this to this with no problem whatsoever. Clean as a whistle, all the detail is there, and there's no pattern noise.

If you can't do this too, it's not the camera that's the problem.

 

[dtaylor]

The Nikon image has pitch black shadows. It's the back sides of the bed, the back side of the desk, and the back side of the curtain. It was over a +5 stop lift, according to the guy who made the images.

They all have some pitch black shadows. But overall the Nikon photo is much closer to your sunflower shot that was given more exposure and the 70D shot. If you shove the brightness on your monitor all the way up, those three files all have quite a bit in their shadows before they are pushed. Your darkest sunflower shot? Nothing but some very dark "blocks" around the locations of sunflowers. (Looks interesting actually, like an old Atari game got scaled up.)

I'm not saying Exmor would have the same noise had it received less exposure. But in my experience you do not want to have to push shadows that hard. Tonality and fine detail are nothing like a proper HDR shot because, as you yourself pointed out, there are very few tonal levels in those last few bits. I generally run into this before I run into noise issues on Canon sensors. I contend that if your darkest sunflower shot was made on a D8x0 that it would have very little shadow noise, but also not look anything like the HDR shots you admire.

What all of these threads are missing is a series of controlled underexposure tests with Canon and Exmor pushed back up to see exactly what one gets at each push (+1, +2, +3, +4, +5, +6). I would be interested in going to a park and shooting that series and providing the RAW files for review, but it would have to wait until a weekend when I can borrow my friend's D800. I know he's using it for paid work for the next couple of weekends.

As for my file, have at it: http://1drv.ms/1vmTXXq

Thank you. I'm getting roughly the same results.