Right. I've been doing some fiddling this weekend, and have confirmed that when equalized for resolution, the 8Mp sensor in my CP8400 outperforms the 8Mp sensor in my E-500 on a per area basis. A 4/3" sensor with 32Mp should outperform an 8Mp 4/3" sensor given similar technologies being applied.
It's a good point and something I find myself doing all too often because I find the myths that DPReview all too often promotes to be frustrating.
Actually, there is significant subjectivity in how the numbers are interpretted from reading the chart. But I agree, it is not rocket science. Objective measurements from the charts show that the resolution increase pretty much exactly as you'd expect given the increase in pixel counts. Noise has nothing to do with any of this.
Yep. It sure is what it is - namely poor interpretations of tests that range from very good to mediocre.
My independant results confirm this. The 50D has right around a 22% advantage at MTF50 and MTF20. At MTF10 (approximately the Rayleigh criteria where all detail extinguishes), the advantage is only about 15%.
Yes, they used to jump on compacts for having low DR, but my experience and tests show the deficit isn't nearly as much as one would expect from all the howling. Further, they conveniently decided not to do DR testing of compacts. And finally, the DR test that they do are often very misleading and result in a lot of confusion.
Yup. I just did a test today to compare the sensor in my CP8400 to the sensor in my E-500. Both are 8Mp CCDs from about the same time period. When images are compared on a "per area" or "per resolution" basis, the CP8400 always provides better results. I even underexposed the CP8400 by two stops at ISO 400 to simulate ISO 1600, and when linear resolutions were equalized and the images were compared on a per area basis, the CP8400 was clearly more pictorially pleasing and seemed to have somewhat better dynamic range.
The Nikon D3 has more than twice the sensor area of the 50D. So it darned well better outperform it on a per pixel and per image basis when it comes to noise.
Don't you see the fallacy in your logic. You are arguing "per pixel" noise. But photography isn't about recording pixels, it is about making images. When comparing two sensors of the same format (50D vs. 40D) the key is to compare full images or equal image areas. All this 100% pixel peeping causes people to compare unequal areas of silicon and hence reach poor conclusions about what whether 15Mp are better than 10Mp.
Assuming similar technology, it is the sensor size that matters. Ultimately, it is a combination of active photon collecting silicon and the efficiency of the microlenses. Larger sensors tend to have more silicon that actually collects photons. That is their fundamental advantage. Modern sensors like the one in the 50D have minor improvements in total silicon and microlense design. So they deliver minor "per picture" improvements in noise and DR, while also delivering more significant improvements in resolution.
Sure. On a "per pixel" basis. But then there are proportionally more pixels. It reminds me of the quote attributed to Lenin, "Quantity has a quality all its own."
Here's a practical example. Same basic exposure. Same area of silicon. Same focal length. Same raw processor. The 8400 image uses four times as many pixels that are 1/4 the size as those used by the E-500. With proper resizing, the 8400 shows more shadow detail and less objectionable noise. Neither is very good, but the more dense 8400 image is better.
The problem is that too many people simply believe what they read wihout putting much effort into questioning it in a critical manner.
I realize that image banding noise does not decrease at the same rate as 2-D random noise when you increase total pixel count in an image. I was simply saying that pixel size is not a direct cause of banding; it comes from sloppy readout on a line-by-line basis.
Low frequency banding will still be confused with signal, and needs some masked pixels to determine that.
Hi Jay,
That the larger sensor always produces lower noise is an assumption not born out by practical observation. My 1DS had considerably more noise at higher ISO than my 40D. Observations are being biased by better electronics and better noise suppression algorithms - it's not as simple as sensor size nor is is all about shot noise. Look at higer ISO noise in medium format sensors. By the logic that larger areas always produce less noise they should be relatively noise free. That decidedliy isn't the case and most manufacturers simply do not try to get serious high ISO from medium format sensors with equivalent pixel densities of smaller sensor dSLR's.
Yes, my point is that there is a point of diminishing returns on pixel density vis a vis noise, just where that point is located on an APS-C sensor is not yet clear. What is clear to me is that published higher ISO frames from the 50D indicate the effects of aggressive noise reduction which significantly impacts detail. Without noise reduction details are diminished by excessive noise as seen in the link I provided as compared with the Nikon D3.
I don't see anything unexpected there. You remind me of characters in the saturday morning cartoons I used to watch, throwing banana peels and ridiculously large objects like anvils and houses into the paths of pursuers. You could have at least made a reference to something specific, there.
I've been thinking for a while now that a very useful test of sensor resolution might go like this.
It has now been demonstrated that it gives close to the theoretical 22%. You may argue that this is not 'significant', but it is all that a 50% rise in pixel count has to offer. Perhaps your view is now that Canon should have gone for a bigger increase?
You are entitled to agree with what you want, but given the patent flaws in DPR's rests, the posting of tests which unambiguously show the contrary and your inability to post any supporting evidence of your own, we may be excused for coming to the conclusion that your agreement is based on faith rather than evidence and reasoning.
What would be the linear adavantage of 15 over 14MP - 3.5%? How likely is this to be observable outside of laboratory conditions? In addition, with its FF sensor, the DSLR/14 will have the advantages of magnifying lens deficiencies 1.6 times less.
No one has ever said such a thing. Of course a very bad big sensor migh end up performing worse than a very good small sensor. The 5D was a case in point, a poor sensor rescued by being big.
MF uses obsolescent sensors designed for science and aerospace. they use read chains built from off the shelf components, because the market doesn't have the volume to sustain the huge development spend of the consumer markets. However, even with obsolete sensors and off the shelf read chains against state of the art components, the extra sensor area brings home an advantage at least at low ISO's.
Once again, you are entitled to a view, but you keep on putting forward opinions without any supporting evidence. It may be that there is a point of diminishing returns, bu where and why?
and based on much more extensive and compelling evidence, many of us will disagree.
Then you wouldn't mind furnishing this "demonstration"??
Bob,
Much more "extensive and compelling evidence"??? It's more compelling because it supports your assertion that "there are patent flaws in DPR's tests"? It's more "extensive"?? You accuse me of making unsupported statements then proceed to make statements you can't apparently support. Hmmm...
