p.1 #2 · For those of you looking for 1.3x in the 20d
I've been saying for some time that the larger sensors are a lot more expensive to manufacture, but no one seems to believe that this difference is significant. If those figures are accurate, it should give the 'full frame for under $1000' crowd a dose of reality. The 1.3x sensor alone costs the same as a Digital Rebel. Add in the weather sealed body, advanced AF processors, and larger viewfinder, and the large difference in cost between the 10D and the 1D Mark II becomes a little easier to swallow, though there is little doubt that Canon is making a lot of money for every Mark II sale. Yet, even if Canon were to take away the advanced AF system and robust body, there is little chance that a 1.3x camera will come out any time soon that costs less than $2000.
What is very interesting is that the 1.0x sensor is not that much more expensive than the 1.3x sensor. Given that the 1-series shares the same body style and AF electronics, it seems clear that Canon makes a killing on 1Ds sales. Not accounting for the R&D costs for the larger sensors (which also trickle down to the lesser cameras like the 1D), the 1Ds probably does not cost THAT much more to manufacture than the 1D Mark II, but Canon makes a LOT more money off it. Granted, that money probably goes a long way toward covering the enormous R&D costs of developing the sensor, but they are clearly milking the 1Ds for all it's worth. I wonder if the 1Ds II will feature a price tag that is lower than the current 1Ds... maybe $6000 MAP? If it comes out next year, 1D prices will probably already be in the $3500 range, so the $6000 will still be a good way to stratify the line while simultaneously making the camera more approachable, and make a convincing case against the upcoming D2x, whatever form it may come in.
A tantalizing thought....
Arka C.
Aug 18, 2004 at 01:55 AM
pidge Offline • • • • [X]
p.1 #3 · For those of you looking for 1.3x in the 20d
Dang.....what bothers me is their yield for the larger size sensors. If my company produced wafers with those yields, we would go bankrupt. Canon really need to go back to the drawing board and improve on this yields. That is ridiculous.
p.1 #4 · For those of you looking for 1.3x in the 20d
Arka: In the original DPR post, there was this paragraph:
If that doesn't look bad enough, I had to reduce the defectivity by a factor of 2 for the Full Frame sensor: otherwise, the model only gave a yield of 1.8% (i.e. no working die per wafer on average). I was sort of happy to do this on the assumption that it translates into many more pixels that don't work on an FF sensor than on the others. Clearly, one might say even more dead pixels are acceptable and ask for further modelling with changed defectivity assumptions, but remember that this number doesn't only relate to dead pixels - the electronics of the sensor has to work, too: if there's a fault that takes out a whole row or column, then the sensor is probably useless. This is actually a pretty rosy view of the cost of an FF sensor - if I'd only reduced the defectivity by 1.5, then the cost is $2745 (yield falls to 6.1%). And the whole notion of repair (which is assumed in this model) is a bit bogus - you can't fix a dead pixel like you can a dead RAM cell by swapping in new lines......Show more →
Without the adjustments she made in her earlier calculations for full-frame, going from 1.3x crop to 1.0x crop (full frame) drops the yield by a factor of 4.5 (from 8.1% to 1.8%) and, on average, results in a 41% chance of there being even a single working die out of the 23 on the wafer. The numbers are pretty scary--I'm fairly certain that the 1Ds is raking in some nice per-unit profits for Canon, but I suspect we have quite a ways yet to go in terms of manufacturing refinement before full frame sensors approach amateur-friendly prices.
Then again, DSLR's seem to be making remarkably large price/performance leaps every couple of years...
p.1 #5 · For those of you looking for 1.3x in the 20d
I never said it was going to be as cheap as the 20D, but I guess I was hoping for something to fll the gap between the 10D and the 1DmkII, both feature and price...
You left off an important bit: Some things to understand - this isn't REAL: the cost model is not up to date with all the latest information
The cost killing is -- as is often the case -- all about yeild. A few years ago, larger LCD manufacturers were running yields down in the single digits (and costs where sky-high). I have not doubt that one of the things Canon is doing well is getting better yields. Not only is the writers information not up to date, Canon's processes are not "average", they are state of the art; cutting edge technology. This can make a *huge* difference to chip cost.
Sorry this information is interesting, particularly from a "scale" point of view (yields of 1.6X relative to FF) but has so many very shaky premises that it stops at interesting.
I have zero question that FF is too expensive for Canon right now, but projecting is extremely dangerous in the chip business.
p.1 #7 · For those of you looking for 1.3x in the 20d
Wow the yields are really low on the larger sensors. This is probably the part Canon is trying real hard to fix. once the yields reach 80-90 percent on the larger ones then you can expect to see 2000-2500 1 series bodies
p.1 #9 · For those of you looking for 1.3x in the 20d
The big issue for me is the buffer and being able to write to the CF while you shoot. If that feature is not improved over the 10D, it will be a non-issue for me .
p.1 #10 · For those of you looking for 1.3x in the 20d
I am not an Fab experts. But our customers are the Fab's, so I'm not completely an outsider either.
I want to say just two things.
One,( Arka ), I agree that there are non-linear cost variances on the chip area.
Two, this is a *Model* as others point that out. I don't understand where his model parameters are coming from. Without disclosing the numbers, I can tell you that the "defectivitiy" in terms of "the number of defects per wafer" is very very low for the production lines. There is a point where it doesn't make economical sense for a Fab to produce the chip. The yield for the FF sensor is way better than the model suggests. Trust me.
p.1 #11 · For those of you looking for 1.3x in the 20d
Yields will never reach 80 -90 percent. They don't reach that on cpu's. They will be very happy to get to thirty or forty percent ... unless you don't mind a 1dx with twenty or fifty dead pixels out of the future twenty million?
p.1 #12 · For those of you looking for 1.3x in the 20d
Pavel wrote:
Yields will never reach 80 -90 percent.
The model says 11.6% or 2 working dies from a 300mm wafer. During R&D or early stage of ramping, the yield can be that low. But for the production line, I would be very surprised if the yield is so low.
I don't know if the yield for the FF sensor will reach 80-90% eventually or not. Some chips yield close to 100% today. CPU has always been the biggest challenge just because it requires the most advanced technology for any given time. The imaging sensors are facing different challenges. They are not yet challenged by shrinking design rule. But the physical die is just big.
Going back to the original issue of how much a FF sensor costs to Canon, I still stand by my position that it will be closer to $100 than to $1000. Am I too naive ?
p.1 #13 · For those of you looking for 1.3x in the 20d
Pondria wrote:
Going back to the original issue of how much a FF sensor costs to Canon, I still stand by my position that it will be closer to $100 than to $1000. Am I too naive ?
Operating costs, I would tend to agree that is is closer to $100 then $1000. But you have to count a lot of NRE costs in there. Plus, I really think that currently Canon might have some trouble with production, though I certainly can pin down *what*. They seem to change AA filters more often then some people change underwear. I suspect that a major problem maybe production of high quality, very consistent AA filters, but I don't *know*. Canon is an extremely opaque company; many you could do some of the basic research and pop up with some good cost estimates, etc. Canon, bah. I am surprised "serious" investors can find enough to put money into the company.
p.1 #14 · For those of you looking for 1.3x in the 20d
Matt Drown wrote:
1.3x crop - 27x18mm - 42 raw die per wafer, 3 working die per wafer (8.1% yield) - cost $920
FF - 36x24mm - 23 raw die per wafer, 2 working die per wafer (11.6% yield) - cost $1373
I'm assuming the difference between the FF and 1.3x yields are due to the density / individual pixel size, where the 1.3x are smaller and therefore you get smaller yields from the lithography process.
If the 1Ds2 is going to have an increased density such that it matches that of the 1D2, there is no way it's going to have better yield than the 1.3x. Who knows what the manufacturing cost increase will be?
I remember Canon announcing last year how they have new fab technologies that will make their CMOS sensors easier/cheaper to manufacture, but given how these are trade secrets and never patented, who knows what the real cost structures and yields are. Given how few 1Ds's they sell every year, I would not be shocked by a 10% yield. The R&D necessary to improve that yield would alone eat up most, if not more than, the profits from the 1Ds so far. IMHO.
p.1 #15 · For those of you looking for 1.3x in the 20d
pidge wrote:
Dang.....what bothers me is their yield for the larger size sensors. If my company produced wafers with those yields, we would go bankrupt. Canon really need to go back to the drawing board and improve on this yields. That is ridiculous.
These are good yields for their sizes. When the 1Ds was first introduced the yield rate was much lower and thus the chips were more expensive. Factor in recovery of R&D costs as well and you'll realize it took awhile for the 1Ds to make any money.
p.1 #16 · For those of you looking for 1.3x in the 20d
Just a quick note, this information is not mine. It's originally from a post that is linked at the start of the post. I just found it interesting because I hear all the complaints about people wanting full frame sensors on the sub $1000 cameras. Please check out the full message on DPReview, it has more information about how the numbers were derived.
In the end this is technology, and about the only thing you can assume with technology is that it will get cheaper, faster, and better over time. So yes, we'll have the potential for cheap FF sensors in the future, we'll also have low noise dense sensors...
