Developments in technology

[Efka76]

Dear colleagues,

I would like to hear your thoughts about developments in photographic technology and reasons why such developments are so slowly implemented. When I look at digital photography and prices it is obvious that the first digital cameras were extremely expensive and not affordable to general public. However, during the last decade there were quite significant developments in digital cameras (increased MP, better DR, ability to shoot with higher ISO). I understand that 5-7 years ago FF cameras were very expensive due to the fact that FF sensors were very expensive. Accordingly, cheaper (crop) sensors were invented in order to make digital cameras more affordable to general public.

I still do not understand why after so many years FF cameras are so expensive as sensor production should be much much cheaper (R&D costs already amortised, mass production is already in place). Also, I can not understand MF camera costs, which rocket to the sky .

Also, there are not so many improvements in Canon's digital cameras technology. As far as I understand in digital camera the most important things are sensor and processor. If you take a look at computer area you can notice very significant improvements in processors speed, their architecture and etc. Canon, for example, still produces 5 year old 7D, which is based practically on outdated technology (I admit that 7D is superb camera, however, it could be even better). So, why is it so hard to photographic companies to try harder and develop better sensors and image processors on a yearly basis. They ear sufficient profits, which could be invested in R&D. Of course, I understand that improvement in optics are very costly and even slight improvement is very hard to achieve. That's why we are still using lenses , which were developed 10y (or even more) ago.

The same applies to MF cameras. I can not justify their costs. In my opinion (maybe I am wrong) the difference between MF and FF is the sensor, processor capabilities and larger lenses. If MF sensors are produced as mass production their costs should be much lower and MF cameras would be more affordable. I have impression that Canon and Nikon are basically milking the same cows for many years and do not try very hard to due to lack of real competition.

Any thoughts on that matters?

 

[dgatwood]

Full-frame cameras are expensive for two reasons: economies of scale, poor fabrication yields, and inefficient use of the wafer.

Economies of scale are straightforward. When you sell more of a product, the cost of tooling up the manufacturing line is amortized over a larger number of units, as is the cost of R&D. Additionally, if you sell more of a product, yield becomes more critical (because if you sell a million units, even a 1% failure rate is a *lot* of returns), which means that you're forced to do more work up front to polish the design, which tends to pay off in the long run.

Fabrication yields are also directly affected by the die size (the chunk of silicon that turns into a single chip). The larger the die size, the more problematic flaws in the silicon wafer become. Suppose you know that 10% of the surface of the chip will be unusable (it really isn't that high—I chose a big number to make the math easier). Suppose that manifests itself as a stripe across the middle of the wafer. If that wafer has a hundred chips on it, that might affect only the ones near the center—maybe 15 chips, which gives you an 85% yield—close to the 90% ideal yield. If that wafer has only four chips on it, that same stripe might affect two of them, resulting in only a 50% yield. If you're really unlucky, it might also affect the corners of the remaining two parts, resulting in a 0% yield.

Of course, realistically, a wafer that bad would be rejected, but there are lots of random flaws that can occur when manufacturing a chip, and the flaws almost all occur in rates that are proportional to the surface area. If you know that you'll have one bad pixel out of every million and a 10 MP crop body sensor, you'd expect 10 bad pixels on average, so if you consider 20 to be the bottom acceptable limit, most of your chips will pass. If you move up to a full-frame sensor with the same density, you're now at a 25.6 bad pixels on average, and most of your chips will fail.

Finally, there's the question of inefficient use of the wafer. The wafer is round, which means there's some waste whenever you make a square chip on it. The bigger the chip, the bigger the waste. Now if you're careful, you might be able to print some other minor parts (e.g. voltage regulators) in the remaining space so that you don't waste as much, but otherwise, there's a significant overhead to bigger dies.