-
Welcome to the new forums! Please read this first. For known issues we are working to resolve, click here.
Full Frame Question
- Thread starter mailman88
- Start date
Is any progress in yield rate and cost unthinkable? I remember that the audio press in the early 80's claimed that CD's would forever be expensive due to the low yield rate. Now we burn them at home...
200 aps-c sensors vs. 20 on a waffer: a factor of ten. Plus bad part of the waffer affect the total yield more than bad spots on a waffer with many aps-c sensors.
Plus the triple exposures stated. Canon is stating that a FF sensor costs 10 to 20 times more than an aps-c sensor.
Those of you asking for a $1000 6MP dSLR - it won't be happening anytime soon. Like I said before a FF camera with few pixels is hardly any cheaper than a FF dSLR with say 12MP+.
The lowest MP count FF sensor likely costs over $1000 itself let alone the body and all its components inside.
100s of millions if not billions of blank CDs are made each year, and even if all dSLRs were replaced with FF sensors, the count would only be around 10 million - that's hardly high volume by semiconductor manufacturering.
nickleback
Forum Pro
It's not yield rate, it's yield. Here's a quick demo.
http://tams-www.informatik.uni-hamburg.de/applets/yield/index.html
Plug in 39, 27, 40 and click "apply selected values". That's a wafer with 35mm sensors (with 3mm of non-imaging border). How many imagers per wafer? That's your yield.
Now plug in 9, 7, 40 and keep clicking. That's a wafer with 1/2.5" sensors. How many imagers per wafer.
Now imagine each processed wafer costs $1000. It could be more, it might be a bit less. How much per imager?
And your analog is very bad, for two reasons.
First, maturity. In the early '80s when CDs first came out, Optical discs had been in development for about 15 years, but the first commercial product (laserdisc) was less than 5 years old. It was new technology. In contrast, image sensors are based on the same processes that make memory chips, CPUs, etc. These processes have been in development for almost 60 years, with commercial applications starting more than 40 years ago, In addition, there have been plenty of financial incentives along the way to develop the ability to make larger chips. Server CPUs are smaller than APS-C imagers. If they could be made larger, they would be.
Second, market size. After CDs were introduced, optical discs became a big market, and CDs made up well over 90% of that market. It's no wonder development was aimed at CD production. In contrast, image sensors are a tiny fraction of IC production (single digits percentage at best), and large image sensors are a tiny fraction of image sensor production.
If you want to believe that large image sensors will be the push needed to develop the ability to make large ICs, then you don't need to believe in the tail wagging the dog, you need to believe in the flea on the tail wagging the dog.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
nickleback
Forum Pro
They haven't yet shown up for the party, and all the booze is already gone.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
Maybe it's the waffer size, but Canon says they get 200 aps-c sensors or 20 FF sensors from their 8 inch waffer. That demo only showed 12 FF or 64 aps-c sensors. But it still shows the dramatic difference in how many sensors can fit at different sizes.
nickleback
Forum Pro
Sometimes companies do sell below cost...
http://en.wikipedia.org/wiki/Tying_%28commerce%29
if they believe they can make it up in other ways. Kodak and Polaroid used this business model, it worked for quite a long time. Canon might be doing this with the 5D, if they believe they can get more high-profit lens sales from 5D buyers. But if they are doing this, that doesn't bode well for folks wanting a camera like the 5D but half the price.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
nickleback
Forum Pro
Yep.
You can fit around 25, but who's counting? But good luck getting more than a few of those to work! Hence the yield is quite low.
I think you'd better check your math. It's a bit more than 2 times. But the yield is much better because while defects might knock out 23 of those 25 35mm sensors, yielding 2 per wafer, defects would knock out 30 of those 50 APS-C sensors, yielding 20 per wafer.
Yes. They'd like to keep their shirts.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
The MF digital market is virtually non existent because there aren't any major player in this field. Until someone with the development resources and who makes their living from selling more than 2000 digital cameras a year, digital MF won't be in the reach for the advanced amateur and hence won't "happen".
I beliveve Pentax can play a role here as the 645D developing cost are to a large extent based on already profitable products; it share the finder, shutter, and chassis from the film 645; basically unchanged from 1984. The camera control electronics and lots of the image processing electronics is shared with the K10D and possibly coming K-series bodies. The sensor will be expensive but camera around it cost nothing even close to what eg. the Nikon D3 body costs. It will also be smaller and lighter and will be a field camera like the film Pentax MF cameras.
nickleback
Forum Pro
Image sensors have a non-imaging border, let's call it 3mm.
35mm sensor: (36+3) (24+3) = 1053mm^2
APS-C sensor: (24+3) (16+3) = 513mm^2.
Please explain how you can fit 10 times 513mm^2 (i.e. 5130mm^2) into 1053mm^2.
The yield difference is 10x. The area difference, and therefore approximately the number of devices you can layout on one wafer, is 2x.
I take it you've never worked near a fab lab. They handle with care and there are still defects. You cannot have a room that's 100% clean, or a blank that is 100% pure.
A blank costs several hundred dollars. Processing costs about $1k. You don't know if you've got good or bad parts until after you've done the processing and have tested.
If consumers will always choose bigger because it is better, then explain what happened to the medium format film market.
(P.S. before about 1950, medium format was the consumer format of choice)
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
nickleback
Forum Pro
It's physically impossible. I've seen the white paper. I work in a technical field and have worked with marketing, so I can understand how this happens. Most likely marketing misinterpreted what engineering told them, and didn't go back to engineering (or the engineers were too busy) to get final approval on the text.
You can do the math yourself, it's really not that hard. It just takes a minute, and you'll know more than Canon marketing. Area of a circle is Pi*r^2. 1 inch = about 25mm, so 8 inches = 200mm. Area of a rectangle is w*h, and imagers typically have at least 3mm border, often more, so...
8 inch wafer: Pi * (100/2)^2 = 31415mm^2
35mm sensor: (36+3) (24+3) = 1053mm^2
APS-H: (28+3) (20+3) = 713mm^2
APS-C: (24+3) (16+3) = 513mm^2
How many of each type can you stuff on an 8 inch wafer? Just divide and round down to the nearest whole number...
35mm: 29
APS-H: 44
APS-C: 61
Note that a wafer is a circle but sensors are rectangular so the theoretical maximum won't be reached. There will be wasted space. The bigger the rectangle, the more space is wasted.
I believe Canon's number of 20 35mm sensors per wafer. It's hard not to, you see it in the picture. I'm guessing the 46 in the paper actually refers to how many APS-C sensors can be laid out. And the number of APS-H sensors, despite being stated in the paper at 46 (which is probably the APS-C number...), is more likely about 32.
But the difference is not 20 vs 32 vs 46, that's just the layout. Actual yield is probably around 2 vs 6 vs 20.
Don't forget the non-imaging border and room to cut. If you use
How many can fit is only half of it. The other part is how many of those ones that fit actually work. That's why they show 2 colors -- good vs defect. That's why the cost goes up exponentially with area, you've got two factors that multiply. You lay out fewer, and a smaller percentage of those will work. If there were 0 defects (and that's a pipie dream), cost vs area would be almost linear (almost because of the rectangle-vs-circle issue).
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
nickleback
Forum Pro
Silicone is jiggly stuff. Used for caulking windows, medical implants, etc.
The paper is talking about Silicon, badly.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
nickleback
Forum Pro
I don't think it's a chicken-vs-egg problem, it's the other end of the continuum -- extinction. 35mm film pushed MF close to the brink (MF was mainstream before the 1950's), digital has it teetering on the edge.
There's much less reason for MF digital to exist than there was for MF film. Cost plays a large part in this. Same reason why APS-C digital has taken over the market 35mm film used to occupy. Quality is good enough and the price is right.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
Clint Sanders
Senior Member
That's how many they can FIT on a waffer.
As Canon said, 20 defects per wafer can ruin the whole crop.
Anastigmat
Forum Pro
Your calculations are wrong. The area of a 35mm sensor is 1053 square mm, not 1053mm^2
A silicon wafer is circular, and an image sensor is rectangular.
"an 8-inch (200mm) wafer has a surface area of approximately 48.7 square inches"
http://www.memc.com/co-as-description.asp
48.7 sq. in. = 31,419.3 sq. mm
20 x 1053 sq. mm = 21,060 sq. mm, or only about 2/3 of the surface area of an 8" wafer are usable when making full frame sensors.
Even though a full frame sensor is only about twice the size of the APS-C sensor, the sensors cannot use the curved border areas of a silicon wafer. When the rectangular sensors are lined up on the round wafer, only about 20 of them will fit, not 30 as is calculated based on surface area.
You are mistaken. The area difference is 10X because you are fitting rectangular sensors on a circular wafer.
I did not say there are no defects, but according to Canon, scratches and dust are the main culprits, not defects on the silicon wafer. With extra care in handling and extra clean rooms, the yield rate can be similar to APS-C sensors. Granted, one defect can wipe out 1 full frame sensor or 1 APS-C sensor, so the failure rate is 1/20 vs. 1/200 for one defect on a silicon wafer. However, 2 defects may both fall within a single full frame sensor but the same 2 defects may be found on 2 different APS-C sensors since they are smaller. So the defect rate is not always linear. 2 defects do not always wipe out twice the number of full frame sensors.
Canon knows. Canon is not blaming failures on bad blanks. Producers of bad blanks will soon go out of business in this highly competitive market. We can pretty much discount defects in the silicon wafer affecting yield.
The medium format film market died because even 35mm full frame sensors can equal medium format film in quality. Medium format digital sensors have replaced medium format film.
Before then, film quality is so bad that anything less than medium format was unacceptable. People picked medium format over larger view cameras because of portability. It is always a balance between image quality and portability. 35mm camera equipment is more portable, and the image quality is acceptable to most people. The price is right too. It appears that the image quality of any film format that is smaller than 35mm, although slightly more portable, is unacceptable. Gone are half frame, 126, 110, disk film and APS film formats. These were heavily promoted to consumers but they ultimately all vanished because they produce prints that are inferior to 35mm. 35mm has not killed medium format film. In fact, many consumers turned to medium format film in the 1990's. 6 x 4.5 format film cameras flourished in the 1990's, with Mamiya, Pentax, and Contax making lots of them and selling them to serious amateurs. These cameras offered higher quality than 35mm film and portability is only slightly worse than 35mm. These cameras would have continued except that digital cameras and backs have made them obsolete due to both high film and processing costs and inferior image quality.
Clint Sanders
Senior Member
Who told you that? Have you seen these clean rooms? How do you propose they be more careful?
Wrong. We can pretty much guarantee defects affect yield. And with huge chip sizes, this affect creates great cost.
nickleback
Forum Pro
square mm is the same as writing mm^2, it just takes more characters.
I'm well aware of that. In fact I pointed that out in another post.
Funny, that's the same number I came up with.
Yep, regtangular peg, round hole. There will be some waste.
Fine.
APS C is on the order of 500mm^2, or half the size of 35mm. So even if you can fill up the entire 1/3 wasted space, that's 20 (number of 35mm imagers) * 2 (APS-C is half the size) * 3/2 (reciprocal of 2/3) = 60. And in fact not even 60, as it is still a rectangular plug in a round hole. There is still going to be some waste. So it will be less than 60. Definitely not 200, that's physically impossible.
Scratches are defects on the wafer. Dust are random things that land on top.
No it won't, larger devices will always have lower yield percentage, and much lower overall yield, because of the compounding effects of size (less laid out on dies) plus defects (more of them are toast). Reducing defects helps, but even if defects are 0 (which will not happen), the yield just drops from exponential to almost linear.
And in fact with large sensors that's the only way you get any yield at all.
Again, you've never worked in a fab lab.
I'd say 35mm sensors have replaced medium format film. The huge drop in MF digital business attests to that. And APS-C sensors have replaced 35mm film. The huge drop in 35mm digital business attests to that.
I see, so Leica cameras in that era produced unacceptable results?
And convienience, and cost.
And medium format did not kill large format. However each were relegated to fairly small niches.
...and the world has gone digital. 35mm digital fills in the niche that MF film once occupied. MF digital now fits in a much smaller niche.
--
Seen in a fortune cookie:
Fear is the darkroom where negatives are developed
john p vansteenberg
Forum Enthusiast
Too many assumptions for the right answer to emerge.
Simplify!
Imagine, IF YOU HAD THE POWER TO MAKE ONE, a digital Rebel ($699 street) . ADD an extra $350 -$400 for the additional cost of the FF CMOS sensor. Make that sensor a 6-8MP.
NOW for near $1K you have the FF digital REBEL Body and you cream every other camera on the market for the consumer market, notwithstanding differering personal preferences for a specific brand.
Next, you need to deduct the production efficiencies of making 1MM units of this single product or more. That drives the cost down maybe 15-20% from the starting point.
Now image your company is focused on holding the leading market share of ALL integrated technologies: capture, printer, ink, etc. Would YOU make a customer for life, particularly if you could cross market with yout printers and other devices? Of course you would.
Forget the 5D, the 1Ds and all of the others taking current advantage of OTHER technologies that are NOT sensor cost issues, per se. (buffer size, A/D transfer speed, write speed, burst rate, etc.) Strip them out of the equation. Focus on the possibilities and the consequences.
--
Van
(A simple note: 4 years ago I brokered a lot of 500 FF CCD sensors (24x36mm) for MF cameras with 6MP 16 bit technology, just as the 12MPs were arriving.) I am pretty clear on the negotiated price of sensors.
Simplify!
Imagine, IF YOU HAD THE POWER TO MAKE ONE, a digital Rebel ($699 street) . ADD an extra $350 -$400 for the additional cost of the FF CMOS sensor. Make that sensor a 6-8MP.
NOW for near $1K you have the FF digital REBEL Body and you cream every other camera on the market for the consumer market, notwithstanding differering personal preferences for a specific brand.
Next, you need to deduct the production efficiencies of making 1MM units of this single product or more. That drives the cost down maybe 15-20% from the starting point.
Now image your company is focused on holding the leading market share of ALL integrated technologies: capture, printer, ink, etc. Would YOU make a customer for life, particularly if you could cross market with yout printers and other devices? Of course you would.
Forget the 5D, the 1Ds and all of the others taking current advantage of OTHER technologies that are NOT sensor cost issues, per se. (buffer size, A/D transfer speed, write speed, burst rate, etc.) Strip them out of the equation. Focus on the possibilities and the consequences.
--
Van
(A simple note: 4 years ago I brokered a lot of 500 FF CCD sensors (24x36mm) for MF cameras with 6MP 16 bit technology, just as the 12MPs were arriving.) I am pretty clear on the negotiated price of sensors.
Peter Nelson
Veteran Member
Yep, I'd take a FF DRebel at $1,000 and use it as my every day camera.
Like I said eairlier I have the 6mp DRebel 300D and the 1Ds Mark II FF 16.6 mp.
My 300D is an excellent and still viable dslr. Making a FF version would be real nice.
I have the L lenses to use on all my Canon cameras and that's very important as well.
visit my web site http://www.flickr.com/photos/artist_eyes/
Remember to click on 'All Sizes' for better viewing.
Artist Eyes
Like I said eairlier I have the 6mp DRebel 300D and the 1Ds Mark II FF 16.6 mp.
My 300D is an excellent and still viable dslr. Making a FF version would be real nice.
I have the L lenses to use on all my Canon cameras and that's very important as well.
--
visit my web site http://www.flickr.com/photos/artist_eyes/
Remember to click on 'All Sizes' for better viewing.
Artist Eyes
Keyboard shortcuts
- f
- Forum