I recently owned a G10 because my wife wanted a good snapshot camera (something more portable than the DSLR), but I soon got rid of it because it was impractical. It had a ridiculous 14.7 MP on a small sensor.
Which you only need to deal with if you shoot RAW. If you shoot JPEGs, you can shoot them at lower resolutions if you want, and the results are much better than JPEGs of that size from older, lower-MP cameras.
14.7MP is not ridiculous, unless you have low IQ standards. A properly sampling camera should
NEVER, EVER give sharp results on a 100 PPI monitor! If the results are that sharp, you need more pixels. It should take 3 pixels exclusive to transition black to white on a sharp B&W transient.
Not only did it have crappy high ISO sensitivity and low dynamic range,
It has more DR and high-ISO performance as good or better than most other P&S cameras before it. The Panasonic LX3 and the Canon G11 are unusual cameras in that they had reduced read noise - older cameras with 10MP sensors don't perform
as well as the G10.
and LOTS of noise even in ISO 80 (!) the file sizes were twice as big as those from my 20D,
That's not the sensor's fault; that is the fault of Canon's firmware department. The G10 is only worthy of 10 bits at ISO 80, and about 1 bit less for each doubling of ISO. That's with linear data. With a lookup table, far less levels can be used. the RAW files can be significantly smaller per MP (this is true of DSLRs also, except Canon DSLRs are worthy of 12 bits linear). The 1Ds3 only needs about 300 levels for the top stop of ISO 100, currently given about 7000 levels. The data could also be stored in such a way that the highest order bits can be separated because they compress extremely well. Current RAW file sizes may be justified for the purists, but they could be much, much smaller for the near-purists (people only conserved with visible purity).
which produced MUCH MUCH better images. For something like that, a 6MP sensor with increased dynamic range, better high ISO, and higher fps would have been a lot more useful.
6MP is a toy. When the pixels get too big, the late stage noise will start interfering with DR at low ISOs.
The issue is not that the IQ is not being well-controlled. It's that it could be made even better if the sensor circuitry were optimized more for, say, dynamic range, color sensitivity, and high ISO sensitivity.
For any given sensor size, they may all be manifestations of the same thing.
And, if that comes at the expense of MP, that would actually be welcome by a certain group of people (and it's not small). That's because for most photos other than, say, landscapes and portraits that may be blown up, we'd rather have higher IQ than more MP.
Most of those people are going to be kicking themselves in the rear in ten years, when their images look pixelated and jagged on their 100MP monitors, with mild-but-large noise grain.
--
John
), but it's sort of ironic that one can manufacture someone else's sensors, while not having sensors of their own that are as good.
