Sensors, Wafers, and Time

[mastertech01]

I am relatively new to digital SLR photography. I have spent a great deal of time over the past 7 years building high performance personal computers. One thing I learned is that processors can vary in quality dependent on the wafers produced during a given cycle of life. Early wafers tended to be poor quality and the longer you waited the better chance of getting a more pure and more overclockable processor. Certain lot numbers would be highly desirable as a result. This held true for chipsets and video processors as well.
I wonder if these variables also hold true for camera sensors and digic processors. Is it likely that you might get a camera that has overall better performance later in the build cycle than if you buy it the second the camera is avaliable. Are there cameras of the same model with much better sensitivity and performance than others simply because of the purity of the given wafer in production?
And if so, is it possible that the camera reviews we read might have turned out better for a given model had they not used the first ones with first generation chips and firmware?

[Dermit]

I work in the semiconductor industry and a lot of what you are saying is true. No two chips will ever be exactly alike (at the atomic level) but all of them have to pass tests to meet a specification. It's just that some might pass with more margin than others. On any given wafer you can have some die pass spec and others fail. The failures can be caused by several factors. Most wafers have a 'sweet spot' where most of the die pass, and also high risk areas where more die fail than average. A silicon wafer processed by a certain fab will cost the same amount whether they put one die on it or 1000 die on it. Whether you get one good die and 999 fail, or 500 good die and 500 fail. This is why larger die like full size sensors become very expensive. It is also why you try to improve the yield any way you can.

Another factor is that sometimes when a company is not getting high enough yield on a wafer lot they often times evaluate the cause and sometimes changing the design/devices they can improve the yields, but the end product is considered the same product. This is why often times waiting for the maturity of a product might yield you a better performer. So, yes, you can have different versions of the same chip out there with some performing better than others.

Usually by the time a product hits the stores all the 'bugs' have been worked out of the system, sometimes this takes a long time. So I would be surprised if you would see any signifacant differences over time, but it does happen. The chips I am designing today probably will not see a commercially available product for 9 months to a year from now, sometimes longer depending on the complexity and how much new technology we incorporate.

[bsmotril]

Unless you're overclocking your camera, I don't think it makes much difference one way or the other. Any variance in sensitivity of the sensor is nulled out when the camera is calibrated during the manufacturing process.

I am relatively new to digital SLR photography. I have spent a great deal of time over the past 7 years building high performance personal computers. One thing I learned is that processors can vary in quality dependent on the wafers produced during a given cycle of life. Early wafers tended to be poor quality and the longer you waited the better chance of getting a more pure and more overclockable processor. Certain lot numbers would be highly desirable as a result. This held true for chipsets and video processors as well.
I wonder if these variables also hold true for camera sensors and digic processors. Is it likely that you might get a camera that has overall better performance later in the build cycle than if you buy it the second the camera is avaliable. Are there cameras of the same model with much better sensitivity and performance than others simply because of the purity of the given wafer in production?
And if so, is it possible that the camera reviews we read might have turned out better for a given model had they not used the first ones with first generation chips and firmware?

[basroil]

Unless you're overclocking your camera, I don't think it makes much difference one way or the other. Any variance in sensitivity of the sensor is nulled out when the camera is calibrated during the manufacturing process.

pretty true, though the problems of hot pixels and such can be hammered out as production continues (*5d*). and what do you think ISO speeds are? lets see, higher iso speed means ramping up voltage in the amplifier... sounds a bit like overclocking to me ;)

for sensors, it does make sense that the older the model is the better chances it works to specs (good for you and canon), but in the case of DIGIC processors, there's so many P&S out there with the same cores that all problems are likely hammered out before they put them in a single EOS series.

[mastertech01]

Great input, thanks!