Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Sigma 70-200 on D40x
- Thread starter epicwelshman
- Start date
- Sort by reaction score
You should re-read Moore's 1965 paper. He addresses transistor density at minimum cost per transistor as well as the implications of cost per transistor.
To not see the relationship of Moore's Law to the cost of integrated circuits...well, that's amazing. How much did you pay for your last computer compared to what you paid for your first one?
He's correct, Moore's law relates transistor count and costs, but not die size. Full frame sensors are more expensive to manufacture than crop sensors and this will remain true in the future. Full frame sensors don't necessarily have a higher transistor count than smaller sensors (12 MP is 12 MP).
Shrinks (i. e. building the same cpu/gpu using a smaller process) is the major driving force behind Moore's law: you can put more cpus on a wafer of a given size, they run cooler and you can usually clock them higher, because the signal paths are shorter. This is not really that relevant for cmos sensors, because you don't shrink them and the transistor count doesn't double every 18 months anymore. At least I don't see 50 MP dslrs in the near future using crop sensors made for the mass market (I'm not talking about digibacks for medium format cameras that cost as much as a car here).
Moore's law is price/performance for CPUs, it's not directly applicable to imaging sensors (if it were, it'd be relative to number of photosites, not physical size.) Imaging sensors are very sensitive to defects on the wafers, and the photosites can't simply be remapped, they have to be in a particular physical location. Larger sensors are more sensitive to defects, so yield per wafer is a significant issue with imaging sensors.
The sensor is the *single highest cost component* in a dSLR, so the cost issue can't be minimized. Secondly, larger sensor are disproportionately more expensive than smaller sensors. Why would a camera manufacturer purposefully make a camera that's significantly less profitable? Why would a camera manufacturer not want to maximize the number of sensors they could get on a wafer?
Full frame digital is the Medium Format of today. Look at the low adoption rate of Canon's full-frame offerings and you'll see that it's a niche product that has a vocal following, lots of "want it" people and not a lot of "bought it" people. APS-C is "good enough" for most people, and isn't likely to go away because it's "good enough" and the costs for the manufacturer go down a lot compared to 35mm.
Full-frame is a temporal anomaly, which is why Olympus bet on 4/3rds (I think they went too small too soon, but longer-term it's probably not going to be a bad move.)
The other main advantage a crop sensor has over a full-frame sensor is higher resolution over a smaller area (detail) without breaking the bank in terms of sensor costs for ultra-high resolution. The offsetting problem is diffraction.
I highly doubt we'll see the manufacturers stop building APS-C and like-sized sensors unless there's a significant breakthrough that makes smaller work even better. Just like film, there's no going back size-wise even though the results are technically superior.
So, you're saying that a 12 megapixel full frame sensor 5 years from now will cost the same that it costs today?
12 MP at what sensor size?
At 1/1.8"? At APS-C? Without mentioning sensor size, there is no way to answer that question. Cheaper, yes, but only up to a certain degree. Certainly the yield correlates clearly with sensor size and that basic math won't change.
The major manufacturers would not sacrifice margins, unless it's absolutely necessary. For entry-level and amateur-level cameras, I don't think it's necessary. If the Nikons and Canons have the choice between making dslrs a bit cheaper by using improvements in manufacturing in their smaller cameras to make them cheaper or increase the sensor size which makes the whole camera bigger and more expensive (larger mirror box, pentaprism, etc.), I think they'll head the same direction they've been heading now: making them cheaper. FF will stay an upgrade for the time being.
Have a look at the resolution: this has stagnated, too, in main stream cameras. The Nikon went from 10 to 12 MP when they've released the D300, ditto for Canon with its 450D and 40D. Nikon's D3 has 12 MP as well and writing is on the wall that this won't be increased substantially at least for mainstream cameras (Canon's 1 Ds series is not main stream `enough'). In the same way, sensor size will not increase for the main stream either. Whether there will be a $1500 FF dslr in three or four years is another question, but that still wouldn't be main stream. And still you'd need very good glass for it (= $$$).
So you're saying that a 12 mp full-frame, 35 mm, image sensor will cost the same to manufacture in 3-5 years as it does today? Maybe you're right. Maybe not. I suspect not, but that's just my opinion.
Yes. As I said, entry level with a DX sensor is reasonable to keep the prices low enough. Other than than, in the $1500-level and up cameras, I'm guessing that the prices of full frame, 35mm sensors will come down to the point where they can still hit their price point and maintain their margin. As with most digital electronics, the tendency is to maintain the price point while performance, whether it be die size, pixel count, transistor count, increases. I think the first 5 or 6 computers I ever owned were all $3000 +/- , while the performance of CPU, RAM, etc increased dramatically. I realize we're talking about larger image sensors with lower yields, but I am still of the opinion that over 3-5 years, the price per sensor will fall, as will performance of non-sensor digital electronics (Expeed will be cheaper, for example) making them available at the same respective price points with suitable margins for the mfgr. If those prices do indeed fall, the mfgrs can indeed keep the price point and make that same camera full frame rather than compromise on a DX sensor. And, I believe that the marketing pressure comes from the fact that selling a D300-equivalent camera that is full-frame instead of 1.5 crop, in 3-5 years, at that $1600 price point will give them a competitive advantage. The D3, even at $5000, has been a huge seller and given Nikon a competitive advantage.
I agree, full frame cameras aren't currently mainstream. I believe that's because of the cost, not performance. I never saw APS film cameras supplant 35mm film cameras. I don't see why, given the option of moving digital cameras from APS sensors to 35 mm sensors at the same price point, people would prefer APS.
I get the impression that arguments against this concept are based on observations of current technology and the current market. I 'm not talking about current technology or the current market. I'm guessing that the technology will advance in the time frame I'm talking about, and that the market will be substantially different. I do see a trend to full frame. Nascent, I agree, but by this fall, we'll have gone from one full frame camera in the market to five in three years. The D3, even at $5000, has been a huge seller. The D700 vs 5D MkII is a market segment that is also generating a huge amount of buzz among photographers. That stuff was WAY off the radar screen 3-5 years ago.
Things just aren't going to stay the same in the technology arena. Look at how digital photography has changed in the last 3-5 years. This whole concept represents my GUESS as to how things will change in the next 3-5 years. You may not agree. That's fine. I could be wrong. Or you could be wrong. But this is the internet and we all have a soapbox.
We're talking about manufacturing- it's not about unit cost, it's about deltas. The delta will stay about the same, so if a 35mm sensor costs 20x more than an APS-C sensor, that's going to be the same. If you can increase your margins by 10% you're a winner, if you decrease your margins by 10% you're a loser. Even if a 35mm sensor were to come down to APS-C prices, APS-C would still be cheaper by a large margin. Using the rule of thumb that a sensor with 2x the area costs 4x as much at the wafer, you can see that larger is always going to be resisted by manufacturing, especially while it's the single biggest component cost- and given that it's area-based, even if full-frame sensors dropped by 25%, APS-C would be even cheaper. Make a million cameras with a $50 sensor price difference and you're making $50,000,000 more in your margins. The delta is not likely to change, yields per wafer are always lower for larger sensors.
Here are Canon's 2006 numbers for sensors per wafer (yield):
APS-C: 200
APS-H: 46
35mm: 20
That means that you get 10 times as many APS-C sensors per wafer. Then you need three exposures for each sensor instead of one- Canon says that a FF sensor is about 20x the cost of an APS-C sensor when it's all done. Now, that's with older steppers, so once you start to amortize the cost of new steppers that can do FF in one exposure you may get costs down another 5x or so.
Even if you had FF sensors as low as $20, an extra $15-$19/unit profit is huge when you're making 100,000 units, and let's face it, FF sensors still cost hundreds of dollars- five years isn't going to help that much.
Nope, that's not exactly what I've said. I've said that it will become cheaper but not a lot. I can imagine that the D300 and 40D may one day have FF sensors (at a price point somewhere in between the two). Improvements in manufacturing will level off at one point.
The second point was that any improvement in the manufacturing process will trickle down to smaller sensors, too. And that small sensors are much more than `good enough already' that it doesn't make sense to increase costs of cameras.
Just think of it this way: if there are x defects per unit area (it's reasonable that this is independent of sensor size, we assume they are produced in the same machine). Full frame sensors have twice the area, so you can produce less than half of the sensors (wafers are circular) and it's twice as likely that there will be a defect. This is a primitive model. In fact, the larger the chip, the lower the yield for more complicated reasons (the mask has to be larger and focussed on a larger area, etc.), but you get the gist.
So yes, I think there will be a full frame `bonus' you need to pay even in five, six years time.
I wouldn't call this main stream, these are price points of semiprofessional bodies already.
No, these comments aren't based on current technology. Although current technology already shows that the main direction is to make things faster and cheaper. The same with computers: cases got cheaper (and more cheaply built, too), etc.
But cameras aren't computers: the megapixel count is levelling off, so are many features (do you need 5 fps on an entry-level camera, I'm sure that this is no problem electronics-wise).
There is too little of an advantage in terms of image quality for consumers to switch to full frame. Even most ambitious amateurs' needs are more than met with crop sensors. But even if you're right, do you think it's worth it taking pictures with lenses that aren't harmonizing fully with crop sensors (in terms of focal length) for at least five years to come? That's half the life of a lens or so.
Manufacturing consumer items is about cost, not performance. I give you $200M to make a camera, you can make 3M FF cameras or 20M APS-C cameras for the consumer market, you have to come up with a FF camera that's almost 7x more profitable to make the same amount of money. You don't get to alter the yields per wafer on your costliest component- so you're constrained in material costs, which change the number of units you can build for your capital investment. Those numbers are made up, but the principle is the same (just more complex,) it's basic manufacturing.
But we did see 35mm supplant 645, 6x6, 6x7, 6x9, 3x4, and 4x5, 5x7 film cameras in the general marketplace. Why? Because it was the smallest size that was "good enough" for most photography. In fact, 110 film was the consumer film for a good while, where 35mm went from consumer to professional once enlarging lenses, film speed and grain got "good enough."
Bigger was always better quality-wise, but smaller was good enough, and to the manufacturers, smaller meant better margins. People in a consumer market don't prefer or not prefer something, they get something that works for them, or they don't. APS-C is "good enough" and cheaper for the manufacturer by a large margin.
I have nothing else to add. I'm not entirely persuaded, but I'm not invested enough in the argument to pursue it. I opted for a full frame camera, not because I believe in the superiority of full frame, but because I believed in the superiority of the D3, which happened to be full frame.
Nevertheless, some of the arguments provided here are illuminating and I appreciate the well-reasoned and polite discourse.
Nevertheless, some of the arguments provided here are illuminating and I appreciate the well-reasoned and polite discourse.