JohnJ80 wrote:
Except that Moore's law presumes you use the density increases to go to smaller die sizes. FF requires the same die size regardless of how many pixels. Therefore, unless defect densities drop dramatically, it is hard to drop the cost fast or much.
I was referring only to the ability of a camera to process 8 FPS or faster of greater-than-8-megapixel frames, not to the cost of the sensor itself. Obviously a full-frame sensor costs more to manufacture than an APS-C or APS-H sensor, and yields are lower. But the cost comes down steadily over time as processes improve. The cost difference between a 1DsMk2 and a 1DMk2N is not solely due to sensor size. A significant part of it is simply profit padding. You want the highest resolution? You pay a premium.
John Black wrote:
Some people here are assuming all these CTO's and Product Managers are equally briefed and coordinated. Yes, there are product roadmaps, but that doesn't stop a CTO or Product Manager or R&D Director from adding their own opinion. And then we try to read between the lines of what somebody says in another country when English is not their first language If you've work in a large global company, then you know how risky that can be.
, no kidding!
Getting all the execs in an multinational company on the same page is like herding cats. They are not trying to be misleading, they just are.
RDKirk wrote:
Mr Iwashita has said before that Canon is a "traditional consensus-driven Japanese company." That would mean the public would never see a difference of internal opinions. The Japanese are far less loosey-goosey about such things than Westerners in general and Americans in particular.
Perhaps but the company I worked for was global and I worked closely with them on several occassions, our last VP was from Japan as well. While the Japanesse were certainly the most rigid of all the regions, they still had a bit of loosey goosey in them. Besides, thats your interpretation of his statement anyway. It could mean their plans are driven by consensus, but opinios may vary (and publically).
moondigger wrote:
I was referring only to the ability of a camera to process 8 FPS or faster of greater-than-8-megapixel frames, not to the cost of the sensor itself. Obviously a full-frame sensor costs more to manufacture than an APS-C or APS-H sensor, and yields are lower. But the cost comes down steadily over time as processes improve. The cost difference between a 1DsMk2 and a 1DMk2N is not solely due to sensor size. A significant part of it is simply profit padding. You want the highest resolution? You pay a premium.
not so if you have these mega die sizes. It is all about getting the defect density down. Historically, that has come down much, much slower than the density has gone up.
The reason that you get more profitable and the yield goes up over time is that you do die shrinks to reduce the die size (getting the same amount in a smaller die size) and the gains that you get from beating the defect density. Yield enhancement on larger die that don't change size, does not parallel what happens in the merchant IC world where you are buying, say, the same size DRAM for a much lower price 6 months later. You do get yield benefits from learning curve etc... but those are smaller (a lot smaller) than what you get from die shrinks.
Die shrinking on a FF sensor is meaningless since it would no longer be a FF sensor then. In point of fact, the "major" cost reduction that was done in DSLRs was going to the crop formats.
Talking to a Phd friend of mine who designed and processed camera sensors for years, the sensor still is coming in at about 55% of the cost of the camera (that's cost, not retail sellign price).
However you look at it, an APS-H sensor based camera will always be able to have higher frame rates than a FF camera - until you hit the limit on how fast the shutter/mirror can cycle. You probably could up the frame rate on a 1Dsm2 to 8fps with the digic 3, but that also means an APS-H based camera could do 10 or 12 fps with exactly the same processing engine.
To put it another way, if you have to trim image size from your full frame sensor to get the fps rate you need, then why would you pay for the FF sensor in the first place? What it comes down to is that technologically AND marketing-wise it doesn't make sense to merge the 2 cameras, because sports shooters and portrait photogs have divergent primary needs.
JohnJ80 wrote:
not so if you have these mega die sizes. It is all about getting the defect density down. Historically, that has come down much, much slower than the density has gone up.
The reason that you get more profitable and the yield goes up over time is that you do die shrinks to reduce the die size (getting the same amount in a smaller die size) and the gains that you get from beating the defect density. Yield enhancement on larger die that don't change size, does not parallel what happens in the merchant IC world where you are buying, say, the same size DRAM for a much lower price 6 months later. You do get yield benefits from learning curve etc... but those are smaller (a lot smaller) than what you get from die shrinks.
Die shrinking on a FF sensor is meaningless since it would no longer be a FF sensor then. In point of fact, the "major" cost reduction that was done in DSLRs was going to the crop formats.
Talking to a Phd friend of mine who designed and processed camera sensors for years, the sensor still is coming in at about 55% of the cost of the camera (that's cost, not retail sellign price).
J. ...Show more →
From reading the infamous white paper from Canon on their full frame sensor technology they mention a very good reason for choosing the APS-H size. It is the largest sensor size that can be made by a Stepper (chip laser etching machine) with one single exposure. I guess that allows for signficantly faster production that a FF sensor requiring 2-3 separate exposures?
On the more aesthetic side of things many readers/writers to FM seem to indicate that they are very happy with the 1.3 crop on the 1DMk2N as the perfect compromise - a bit more reach but not too significant reduction in wide-angle, good speed capability etc.
On the downside it does prevent access to the increasingly useful range of EF-S lenses which is perhaps not that big an issue for those who will budget for L-glass anyhow?
Having started down this route with FF, APS-C and APS-H sensors perhaps each has their market and Canon dont want to risk cutting out of a market segment without very thorough understanding of what the risks are.
John, I understand what you're saying about die size and manufacturing, etc. When I mentioned Moore's law it was only in reference to the CPU and processing power of the camera.
Talking to a Phd friend of mine who designed and processed camera sensors for years, the sensor still is coming in at about 55% of the cost of the camera (that's cost, not retail sellign price).
We know for a fact that Canon can profitably sell a full-frame camera for well under $3000, given the current double rebate price of a 5D. Heck, after Dell's double discount day a couple weeks ago we discovered that the 5D could be profitably sold for under $2000.
What I'm suggesting is that given improvements in CPU power and memory, 8 FPS or faster could be achieved with full-frame sensors in the 12-13 megapixel range. I further believe that such a camera could be based on the 5D sensor, and that it could be very profitably sold for a retail price in the same ballpark as the current 1DMk2n. I'm not saying the super-high megapixel 1Ds-type camera would go away, only that the 1.3x crop camera would go away. Presumably the next 1Ds style camera would be 3-5 FPS and around 20-22 megapixels.
timbop wrote:
However you look at it, an APS-H sensor based camera will always be able to have higher frame rates than a FF camera - until you hit the limit on how fast the shutter/mirror can cycle.
Off the top of my head, the fastest EOS body ever was the 1D RS -- it achieved 10 FPS but it used a pellicle mirror that didn't have to move. The image was projected through the semi-transparent mirror, at the expense of 2/3 stop of light. Given that precedent, it seems likely that 8 FPS is the design limit for a camera with a moving mirror.
You probably could up the frame rate on a 1Dsm2 to 8fps with the digic 3, but that also means an APS-H based camera could do 10 or 12 fps with exactly the same processing engine.
The ability to process frames has nothing to do with physical sensor size -- 1.3x is irrelevant to the question. It has to do with the sheer number of pixels. If Canon wants to maintain the good noise properties of the 1D series cameras, but up the resolution, they'll likely have to go to a full-frame sensor. The 5D sensor has the same size pixels (and therefore the same light-gathering ability) as the 1DMk2n. Stick that sensor into an 8 FPS body with a redesigned mirror/shutter and you now have a 1DMk3 with full-frame and 8 FPS. Who wouldn't like that?
What it comes down to is that technologically AND marketing-wise it doesn't make sense to merge the 2 cameras, because sports shooters and portrait photogs have divergent primary needs.
I'm not saying they should merge the two cameras; I'm saying they should eliminate the 1.3X crop and go totally full-frame in the 1 series. You would still have a choice between lower resolution (12-16 megapixels) high-frame rate and high resolution (20-22 megapixels) lower frame rate. It's just that both would be full-frame.
I believe the question was merging the two cameras not dropping the 1.3 crop. If Canon would maintain two bodies, I would agree with everything you have said. However, as I said earlier, PJs and sports shooters do not list FF as their most wanted in the next Canon camera. Higher ISO performance, updates to a rather old AF system, reduced weight seem to top the list not FF.
As I recall the big question a year or two ago when one of the interviews was published was not about the particular cameras, but about whether Canon would eliminate the 1.3x crop leaving only 1.6x (EF-S) and full-frame. My comments are based on that speculation.
As for what PJs and sports shooters want, I would think they'd like more megapixels without a low-light noise penalty. The full-frame would give them that, assuming all else (frame rates, etc) could be kept constant.
moondigger wrote:
John, I understand what you're saying about die size and manufacturing, etc. When I mentioned Moore's law it was only in reference to the CPU and processing power of the camera.
We know for a fact that Canon can profitably sell a full-frame camera for well under $3000, given the current double rebate price of a 5D. Heck, after Dell's double discount day a couple weeks ago we discovered that the 5D could be profitably sold for under $2000.
What I'm suggesting is that given improvements in CPU power and memory, 8 FPS or faster could be achieved with full-frame sensors in the 12-13 megapixel range. I further believe that such a camera could be based on the 5D sensor, and that it could be very profitably sold for a retail price in the same ballpark as the current 1DMk2n. I'm not saying the super-high megapixel 1Ds-type camera would go away, only that the 1.3x crop camera would go away. Presumably the next 1Ds style camera would be 3-5 FPS and around 20-22 megapixels....Show more →
I'm agreeing with this.
Doesn't Nikon have the option to reduce the file size and essentially turn it into a smaller sensor and then pick up the frame rate? I'm wondering that would not be a great compromise - 8MP at 8fps or 13MP at 5fps (or even 3).
moondigger wrote:
What I'm suggesting is that given improvements in CPU power and memory, 8 FPS or faster could be achieved with full-frame sensors in the 12-13 megapixel range. I further believe that such a camera could be based on the 5D sensor, and that it could be very profitably sold for a retail price in the same ballpark as the current 1DMk2n.
Yep, that's certainly the most sensible thing to expect. Now that they've tested their 'cheap'/high margin FF sensor with the 5D, there's no reason why they shouldn't put it in the next 1D camera.
MPerdomo wrote:
Maybe this will teach everyone not to treat Chuck Westfall's word as gospel
Are you kidding? In some ways, Canon users are worse than 16 year old readers of tabloid magazines...rumors are a dime a dozen. Just look at all the speculation threads and the "key players" in each one of those threads. It's like those people live to make guesses about Canon's next move/product offering. If there is ANYTHING they can use as material to make further guesses, you can bet that they'll quote it and make it support their prediction.
"Actually, the second interview with the exec backs up chuck's comments. He also addresses the first who seemed to indicate that a merge would happen (simple 'yes' answer to a question that was two parts, non native english speakers). So we have:"
There's been more posted online interviews out there backing the one body comment up. Chuck works for Canon and he's going parrot the corporate line if he wishes to keep his job. If you've followed this stuff over the many years, you would know the truth of what I write in regard to Chuck's ability to comment in regard to product releases (non-disclosure) and what really happens.
You're welcome to believe anything you wish, cause come PMA or when ever Canon Execs, not Chuck, decide to announce, they will.
The ability to process frames has nothing to do with physical sensor size -- 1.3x is irrelevant to the question. It has to do with the sheer number of pixels.
Yes, exactly. A FF sensor has a much larger area than a 1.3 crop camera, and assuming a roughly similar pixel density the FF camera's bus has to push a lot more pixels.
I do see your point about mirror movement rates, but canon is not precluded from putting an electronic shutter back into the 1 series (a la 1D).
moondigger wrote:
As for what PJs and sports shooters want, I would think they'd like more megapixels without a low-light noise penalty. The full-frame would give them that, assuming all else (frame rates, etc) could be kept constant.
But not everyone is as enamored with full frame as you seem to be, and the current IIN has more than enough pixels(and resulting file size) for its intended use. As many others have mentioned, I'd prefer the 1.3X stay, and I'd like to see improvements to AF, dynamic range, battery system and quality control.
Given how carefully Westfall parses his words, I think he could have a future in law or politics... :-)
My hunch is still that at some point we'll see something like the Rebel XTi as the high end crop sensor camera, and that we'll see the niche currently filled by the crop sensor 30D be filled by a full frame camera that is initially somewhat more expensive than a 30D type camera and less expensive than the current 5D, but whose price will eventually drop to a point that is competitive with current high end crop sensor models.
If you've work in a large global company, then you know how risky that can be.
, no kidding!