There may not be anything revolutionary about EXMOR but it is better than what canon has today... Then there are patent issues we don't know about. Canon may be able to make an equal sensor but then why didn't they?
I suspect there was a quality control/yield issue. Yes, the sensor worked, but not enough of them were coming off the production line that met their standards - too many rejects in each batch.
Unless you have evidence, there is little to suggest this is the case. If your premise was that there was nothing special about the D800 exmor then why would canon not be able to manufacture it with similar yields to Nikon/Sony? This seems contradicting to your prior point.
I did say this was pure conjecture, right? :-)
I don't think the Sony senor is revolutionary because it is not based any break through in physics or CMOS design. It uses the same basic materials in the same layers and circuitry as before. (silicon, gallium oxide, etc) It is evolutionary
- a dense packing of sensors over a large area. That's why I don't think there in anything special about making one, other than manufacturing technique.
(for example, revolutionary would be using buckey ball nano tech coated with room temp superconductors :-P )
Increasing the size of a chip or the density of devices on it is always difficult in semi-conductor manufacture. Both factors multiply to increase the likelihood of failure or at lest than ideal performance. The cost of a chip with a new design is never a function of the materials or the process to make the actual chip being sold. It is measure of how many chips were rejected to get one that works.
Perfecting a semi-conductor production line for higher yields is as much an art as a science. Just because it is not new science, there is no reason to believe that every manufacturer can do it. AMD struggled for years with Athlon chip yields, where for Intel, fabricating CPU's with equivalent complexity seemed effortless. Likewise, Nvidia vs ATI on graphics chips. Or it could just be bad luck. A fire. The Thai floods. Contaminants. It has happened to others.
Nobody gets it perfectly right the first time, and sometimes, companies only break even or take a loss on the runs in the early years to be first to market with a performance boost. As I mentioned, Sony did this with the cell processors for the PS1 and PS II and Intel has done it repeatedly over the years with various Pentiums.
Because large, high density CMOS chips have low yields, Sony may have gambled that with big enough production runs, they could produce enough chips that meet the 36 MP, FF Nikon spec. One percent of 1,000,000 chips attempts is still 10,000 good sensors. And there may be enough market for the various types of rejects to offset the cost of some of the discards.
It is not a rip off in any way for Sony to charge a premium to Nikon for the "cream of the crop". Sony says: we're planning to attempt 1 million chips. Out of that, 5,000 will meet your strictest criteria, and another 5,000 will come really really close. Nikon replies: rather than attempting 1 million chips, we'll pay the extra cost for 10 million chip attempts so we can have our pick of the best, and you're free to use any we pass on it any way you want.
That's why I don't think Sony necessarily
got any better yields than Canon - with a much larger semi conductor business, Sony may have been willing to float much larger runs at the same low yield as Canon.
I am sure Canon knew about the Sony chip years ago - the may even have considered using it. They certainly saw the specs, if not the chip itself.
Why do I think the Nikon price caught Canon off guard? Retail price is one of the last things a company set when releasing a new product. There are cost targets during design and manufacture, but actual retail price is set very late. The fact that the D800 preceded the 5DMkIII by a few weeks didn't lessen the surprise, and at that point, Canon's promotional material and product channels were already set up for $3500. It takes a little while to turn a ship that big. Nikon may have even faked Canon out by preparing two sets of channel materials - one with the higher price in the weeks leading up to release, and one with the real price, released at the last minute. It's been known to happen. But I do agree with you, there is a bit of Canon hubris in that $3500 price tag.
I honestly think that talk about said cheaper canon FF camera is just a bunch of nonsense in reaction to Nikon's alleged D600, which still just a rumor. The next canon FF body will likely be a big MP 1Ds4 to compete with nikon's upcoming D4X.
Maybe. I think the next Nikon FF will have fewer MP's (18-24) and will be made from the same production runs as the 36 MP sensor, and will therefore exhibit the same DR, noise and color capabilities.
Just because you can't get 36 perfect MP's with low noise and high DR, doesn't mean the chip isn't good. The pixels of many 10-12 MP chips are not actually 10-12 million single sensors, but grouped sets of 18-24 million sensors, which taken separately, have less than ideal performance, but taken together, produce very good results. One pixel in the group may give you great DR info, another, the best noise floor. Taken in combination, they produce perfect, if lower resolution results.
So a rejected 36 MP FF sensor may make a perfect 18 MP FF. And the yields will be much higher, and therefore the cost of the sensor much lower because you don't need 36 million perfect sensors.
If Canon releases a FF 18-20 MP at substantially less than the 5DMkIII (with fewer bells and whistles of course) then it would be a good indication that they are on their way to a high MP FF like the Sony, and these these are the rejects from that line.
In any case, I think Canon is working extremely hard to get it's high MP, high DR senor production to scale.