PDA

View Full Version : Camera Sensor Chart



D70FAN
11-10-2005, 02:06 PM
Just so people know what I base some of my conclusions on I am attaching a chart that combines all of the different sensors with their x and y dimensions and area, and the relative (not exact) pixel areas in microns for a given average density. The reason these are relative is that actual pixel densities are generally not exactly 2,3,4...16 MegaPixles so these are rounded off a little.

CMOS has a higher noise factor than CCD and that is factored in to the chart. CCD is 2-3 times the cost of CMOS in large die sizes which is why everyone wants to use CMOS for larger arrays. Very few of the consumer cams use CMOS (even Canon) due to the noise difference.

Anyway I hope this helps explain why certain cameras have certain limits due to sensor size, and type.

coldrain
11-11-2005, 03:28 AM
CMOS used to have higher noise, but this is not true anymore for a number of years now. I do not know why you keep that "fact" going around still, you should knoe better. And saying the Canon DSLR CMOS sensors have more noise but just a very good noise reduction processor is not true either. Canon, and lately Sony, have proven that CMOS does not have to have higher noise levels. That the D200 has a CCD has more to do to what is available to Nikon than to what is better. Sony's 10mp CMOS is very nice, but too small for Nikon's DSLR's (not 1.5 crop). Sony is in a transition, they can not just make any size CMOS for a client like Nikon without having the prices rise too much. Nikon has demonstrated that well with their great D2X with new generation Sony CMOS, CMOS has come a long way. The CMOS noise you refer to is just not about the modern CMOS anymore, not since the Canon EOS D30 at least.

So get off your hobby horse and accept that CMOS (as CCD) has come along way the last few years. Old dogmas should be allowed to die a peaceful death.

You can not seriously think Canon and Nikon (and Sony) somehow magically filter noise away yet keep all the detail that was hidden underneath. And you could also look at the fact that for instance the 6mp KM 7/5D and Nikon D70 with Sony CCD have more noise than the Canon EOS 6mp 300D and 8mp with Canon CMOS.

D70FAN
11-11-2005, 06:52 AM
CMOS used to have higher noise, but this is not true anymore for a number of years now. I do not know why you keep that "fact" going around still, you should knoe better. And saying the Canon DSLR CMOS sensors have more noise but just a very good noise reduction processor is not true either. Canon, and lately Sony, have proven that CMOS does not have to have higher noise levels. That the D200 has a CCD has more to do to what is available to Nikon than to what is better. Sony's 10mp CMOS is very nice, but too small for Nikon's DSLR's (not 1.5 crop). Sony is in a transition, they can not just make any size CMOS for a client like Nikon without having the prices rise too much. Nikon has demonstrated that well with their great D2X with new generation Sony CMOS, CMOS has come a long way. The CMOS noise you refer to is just not about the modern CMOS anymore, not since the Canon EOS D30 at least.

So get off your hobby horse and accept that CMOS (as CCD) has come along way the last few years. Old dogmas should be allowed to die a peaceful death.

You can not seriously think Canon and Nikon (and Sony) somehow magically filter noise away yet keep all the detail that was hidden underneath. And you could also look at the fact that for instance the 6mp KM 7/5D and Nikon D70 with Sony CCD have more noise than the Canon EOS 6mp 300D and 8mp with Canon CMOS.

Thanks for the lecture. I am assuming that your expertise in semiconductors comes from years of experience, or at least a fair amount of research.

You may be correct in your assumptions, just the way I may be correct in mine.

I am taking the information I have and drawing conclusions. If you would really look at the trends and results carefully you might draw the same conclusions. Or maybe not.

I posted the graph to try and help explain size and density limitations. Use it or don't. If you have additonal published facts then I would definately be interested and would probably change my assumptions.

Esoterra
11-11-2005, 07:40 AM
T.O in the house!!!!

Classy reply George. Thanks for the info.

Knoe?

TheObiJuan
11-11-2005, 09:05 PM
Microlenses, heat control, and onboard noise cancelation make CMOS chips respond very well to noise.
As coldrain stated, detail is retained and noise is minimized on the 1.6 bodies.
On charts at dpreview showing actual noise, the nikons usually have less, but it is more apparent in the pictures.

D70FAN
11-11-2005, 10:18 PM
Microlenses, heat control, and onboard noise cancelation make CMOS chips respond very well to noise.
As coldrain stated, detail is retained and noise is minimized on the 1.6 bodies.
On charts at dpreview showing actual noise, the nikons usually have less, but it is more apparent in the pictures.

Sorry Juan. 1.6X is at it's practical limit at 8MP. If CMOS is all they have then Canon is bound to move to 1.3x in the future. this is not a bad thing. Canon presented the 5D as full frame to leave 1.3x open for prosumer use.

But if you have published data. I'm more than willing to change my mind. I have seen/read no evidence that CMOS has improved enough to meet or exceed NMOS CCD in noise imunity.

Keep in mind that I am not promoting either side here, but just presenting a chart. You can draw your own conclusions by what is happening in the marketplace.

coldrain
11-11-2005, 10:49 PM
Call me stupid but if 8mp is the limmit, then what is Sony doing with a 10 mp 1.7x in the R1? And with the 12.X 1.5x in the Nikon D2X?

Vich
11-11-2005, 11:43 PM
Call me stupid but if 8mp is the limmit, then what is Sony doing with a 10 mp 1.7x in the R1? And with the 12.X 1.5x in the Nikon D2X?
Maybe exceeding it. But who knows what technical hurdles have been newly leaped by those engineers. Saying that's the limit for all time would be about like saying the Intel x86 chip maxed out at 800 MHz.

Here's one reference showing some of the issues;
http://www.dalsa.com/shared/content/pdfs/CCD_vs_CMOS_Litwiller_2005.pdf

For sure; CMOS used to be notably lower image quality. They (Canon, Rockwell, etc) took the problems one by one. Maybe they even abandoned the traditional "CMOS" and secretly have some sort of hybrid (one early proposal, see 2001 reference here: http://www.noao.edu/meetings/lsst/kozlowski.pdf).

Some of the solutions I've heard about only allowed CMOS to be less bad (lenses over the pixels, smaller ICs), while some could potentially popel them ahead of CCD (like pre-shot compensation values to adjust for hot / dull pixels and other post-processors to reduce noise).

CCD hasn't just been sitting around. According to the above article, they've also addressed some of their inherent issues (power, heat).

Where does it stand today, I'de sure like to know.


George: Nice chart! Maybe the very slight adjustment for CMOS caused some contraversy, but its a very useful illustration of the sensor / pixel size vs. noise relationship. THANKS! Job well done!

D70FAN
11-12-2005, 05:39 AM
Sorry guys. I should have said current practical limit. Meaning that designers using CMOS have to come up with ever increasing patches and work-arounds to compensate for CMOS sensors. This is why Canon is using 1.3X and Full-frame for their professional cameras.

Not a bad thing.

TheObiJuan
11-12-2005, 12:53 PM
I still don't get why Nikon is using a 12.x 1.5x crop CMOS chip on it's flagship pro body.

I love 1.6x and don't care for more pixels, I want better pixels and better noise performance.
I really don't care if Canon upgrades the MP's on a 1.6, actually, I'd rather they didn't.

George, I didn't get around to thanking you for your chart. Thanks for compiling the info and making it intuitive and presentable.