[jeffkrol]

With infinite funding? No. I really don't think so. The CCD creates too much noise and I think that would effect the total DR. I think if the same infinite funding was put into a CMOS, the CMOS could produce more. Plain and simple, it's a more efficient sensor.

Possibly, the answer may already be here. Have a look at whatever camera has the undeniably most DR - what is it using? Compare that to whatever camera is getting the most DR from a CMOS. How close are they? CMOS is a young technology in this respect and will likely benefit from future development in larger increments.

Thought it depended on well size. Which, in a defined area, will always favor CCD. And CMOS was always inherently noisier than CCD.
http://www.dalsa.com/shared/content/pdfs/CCD_vs_CMOS_Litwiller_2005.pdf
Of course this states CMOS is better for DR.
CMOS vs. CCD: Changing Technology to Suit HDTV Broadcast
So there you have it. A non-answer.... (Any small change could easily favor one over the other)

[smile-123]

I haven't read anything from Pentax about how they are using the CMOS sensor, but I have a technical background in them from way, way back in the early days when they were making these devices literally one at a time. I did see that the k20D will have a 14 bit A/D instead of the 21 bit A/D in the K10D and that let's be jump to some conclusions.

From the A/D side of the CCD, ' dynamic range' is simply the range of light intensity that can be registered by the device from a single bit on to all bits on, and that's a pretty wide range when you consider they are using a 21bit A/D. I can't convert that to lumens or anything, but its a big range. Here's where the problem comes in. In order to reduce that to reasonable ranges, in the case modeled after ISO film speeds, the processing engine selects a sub-range of that data and essentially discards all the other information above and below that range. You are afterall, going from a numerical range of 2**21-1 to 2**8-1 (the ** here represents an exponent, i.e. 2 to the 21 power). But, the CCD actually sees the entire range of light all the time. The CCD sensitivity is very noisy under low light (small numbers) plus A/D conversion adds noise, then there's noise introduce by limiting the range of A/D output, and again by conversion in the processing engine.


The case of A CMOS sensor is a little more complex because the senitivity of the device can be adjusted with its bias voltage. Need more sensitivity? Just increase the bias voltage which also helps to keep the device from operating in its noisy range. You no longer have to accept the entire physical range of the device in order to extract the data that is appropriate for any desired ISO speed which allows use of lower resolution A/D converters which give less noisy, more linear results, and all this requires less post processing (i.e. noise reduction in the engine).'

In theory, the CMOS should be a far superior sensor that produces far less noise.

WARNING: I am an old fart and its been a long, long time since I studied this stuff.

[Rickster]

These arguments about dynamic range are interesting. I wonder how many of us actually have systems that can fully display the dynamic range of the current crop of DSLRs let alone the K20D? The average dynamic range of current LCD monitors are between 450:1 and 5000:1 , unless of course you happen to have this puppy ...

DR37-P - Extreme Dynamic Range LED Display - Dolby

Quote:

"The DR37-P is a spectacular breakthrough in display technology. It uses an array of individually modulated LED backlights to provide 10 times the brightness and 100 times the contrast of existing television and computer monitors. The Extreme Dynamic Range display delivers more vibrant images, enabling you to see information onscreen in vivid detail.

* Extreme Dynamic Range
* Over 3,000 cd/m2 brightness
* 0.015 cd/m2 black level
* Contrast ratio > 200,000:1
* High-definition 1,920 x 1,080
* 37-inch screen
* 16 bits per color
* Array of individually modulated LED (IMLED) backlights"

Oh yeah, if you want one you can write a check for about $50,000. That should just about cover it.

[infosyn]

The case of a CMOS sensor is a little more complex because the sensitivity of the device can be adjusted with its bias voltage. Need more sensitivity? Just increase the bias voltage which also helps to keep the device from operating in its noisy range. You no longer have to accept the entire physical range of the device in order to extract the data that is appropriate for any desired ISO speed which allows use of lower resolution A/D converters which give less noisy, more linear results, and all this requires less post processing (i.e. noise reduction in the engine)... In theory, the CMOS should be a far superior sensor that produces far less noise.

Is this possibly the way the K20D Dynamic Range Enhancement feature works? Seems like it would be theoretically possible to convert K20D's CMOS analog values to 21 bits of digital (as in K10), then "tone map" those 21-bit values into a 12bit RAW file using something a lot smarter than a global linear correspondence. Perhaps it could be along the lines of HDR, where you can rescue information from the shadows and highlights by emphasizing those regions. Of course this is a whole lot more "post" processing than the sensor chip does now...

Are there any CMOS sensors which are able to do multiple A/D conversions, possibly while changing the bias voltage, while the shutter is still open, thereby creating more than one RAW file?

[dosdan]

Smile-123, are you familiar with the Clarkvision.com home website?