CCD vs. CMOS

CCD VS CMOS is a bit of the Digital VS Film. If you have the time and want lower power then CMOS wins hands down (Hubble space Telescope, JPL Imaging etc etc.) If you need speed and can afford the extra noise generated by heat and extra power consumption then CCD has it. Both problems will eventually be solved with additional circutry. However I'm sure what is of interest to most of us on this forum is Will the image quality on the 1D be better (significantly) or not. If Not then you have to judge whether the AF and faster FPS rate is worth your while investing in this camera (for some sports and PJ's it might be) or whether to hold on to your D30's for a while longer yet to see what transpires.
So far I have been unimpressed by sample pix but I am impressed by the spec of this camera.
Hopefully someone will post some real samples
Cheers all
-r

[GabeSchaffer]

Re: focusing sensors, the 1D uses a CMOS sensor at the bottom of the mirror box for focusing. I recall reading that it has 10,724 cells on it. This is very different from the CCD imager used to read the image.

Re: softness, digital cameras use a low-pass (AKA anti-aliasing) filter to keep any high spatial frequencies (i.e. sharp detail) from hitting the sensor. This is to prevent detail from hitting one pixel but not an adjacent one, which would otherwise result in stair-stepping and color artifacts.

Naturally, if you're going to have a filter which eliminates any sharp detail, the resulting pictures will be soft. Thus, sharpening is always required for digital images, though some cameras do it (or over-do it, as the case may be) for you, while others leave it up to you to decide if sharpening is necessary.

Note that Kodak makes the filter optional on its pro cameras, but you're not liable to find many people who don't use it (unless they broke it or couldn't afford it in the first place).

[Noel_Carboni]

Phred,

A digital camera of the modern variety measures the intensity of only one color range of light (e.g., Reds, Blues, Greens) at each imager cell, owing to tiny filters mounted over each. The arrangement of the colors, Red-Green-Red-Green-etc. on one row and Green-Blue-Green-Blue-etc. on the next, is called a Bayer pattern.

Thus, the raw data delivered from the imager really looks nothing at all like a full-color image. The camera has to do some heroic processing, including color/luminance separation and color interpolation, to create a final image, and to give anything near pixel-level detail (Zoombrowser/Bibble export this processing to your computer when you shoot Raw).

Given these facts, digital cameras such as the D30, 1D, etc., deliver greater luminance resolution than color resolution. Look very closely at what should be a sharp color transition in an image and you can see this plainly.

Most manufacturers have added a digital sharpening step to overcome the inherent softness the above processing delivers. Canon has been less aggressive about doing this.

Hope this helps you to understand why images aren't naturally pixel-sharp.

-Noel

[ October 02, 2001: Message edited by: Noel Carboni ]

[BradleyPhillip]

Most "sharpness" information is carried in the luma channel, rather than the chroma channels, and is generally thought of as the slope of a contrast change. (ie. If you could go from very very dark in one pixel to very very light in the next pixel, the picture would contain a very 'sharp' transition).

The problem with digital imagers like CCD and CMOS sensors, scanners, etc., is that each pixel can only see an average value. Typically a high contrast transition will fall on a pixel, and the pixel will see the transition as neither light nor dark. This 'grey zone' tends to 'spread out' the transition compared to how it would be rendered on film, and is a major source of the 'softness' of digital images.

Other forms of filtering (anti-aliasing, false colour) also introduce additional softness into digital images, as is reduced chroma resolution, as described by Noel.

In camera sharpening can be perfomed my most, if not all cameras to 'hide' this fact, by artifically increasing the slope of contrast transitions in the luminence channel.

It's a boon to digital photographers to effectively turn this off, because the artifical sharpening *is* different from the original image, and contains its own artifacts. When the image further processed by the photographer, these artifacts (which were only introduced after the image was captured) also become enhanced, and lower the image's quality.

bradley phillip

[Uwe_Steinmueller]

>Re: softness, digital cameras use a low-pass (AKA anti-aliasing) filter

Not all. with the 760 you have the choice and my choice is without AA.

Uwe (www.outbackphoto.com)

[JohnRay]

quote:

---

Noel Carboni wrote:

Regarding "noise reduction", Canon's soft images (more so in the D30, but in the 1D as well as compared with the competition) are solidly on the track of "less is better". They're not soft because of noise reduction. On the contrary, they're soft because they haven't been oversharpened/overprocessed by the camera!

---

Right on the mark Noel. This is the main reason I purchased the D30 (along with its very low noise). I had been looking for something to replace film for most work up to 11x14" and the D30 was finally a camera that did not assume everyone was going to make 5x7 prints and instead elected to leave the data alone. It also makes the image processing faster.

RAW D30 images will upsample to several times their original size without the introduction of noticable noise, and because they are not sharpened by default, the size of sharpening artifacts are not increased in size in the process. In addition, sharpening algorithms work best when there are a larger number of pixels to sample against the edges, so yet another bonus. If the image is heavely sharpened by the camera before upsampling, the sharpening artifacts become far more noticable, and yet the overall image still seems to become softer to a degree. Applying the correct amount of additional sharpening for print at this point is all but futile.

quote:

---

Phred wrote:

Ok, please share the information you have regarding what effect noise reduction, and while we are on the topic, any other engineering issues (pixel size/shape, color masks/filter, color interpolation) have on raw unprocessed image quality (resolution, sharpness).

---

Forget all the crap about pixel size, shape, masks, etc. The reason the D30 sensor is so noise free is partly due to the fact they have the gain amplifiers and A/D converters directly attached at each sensor. This not only removes the amount of extranoius noise that can be picked up by PCB (Printed Circuit Board) traces post sensor, and pre A/D converters, but it also increases the signal to noise ratio of the digital signal by virtue of this. This allows lower voltage devices and buffers to be used to transfer data. Once it's digital it takes a lot to screw the data up if it's clean.

quote:

---

Ahhh, we are still having this chicken and egg discussion. Post processing and sharpening of course are only needed because of a pre-existing condition. One doesn’t have an infection because they haven’t taken antibiotics..

---

Hmmm... I have to think about this for a second... Nope, sorry, you're wrong (IMO of course).

The only pre-existing condition is the one that 'ALL' digital camera images have. The only difference is that most digital cameras pre-sharpen the image before it is written to the CF card or MicroDrive, and the D30 does not for the most part. This lack of 'in camera' sharpening allows you to add the amount, and type, of sharpening to the areas you wish to sharpen, post saving the data file for the size output you intend to use. This is a 'BIG' deal (I wish I could have used a larger font for BIG), and it's an important distinction between the D30 and other digital camera offerings more than a year after the introduction of the D30.

Let me ask you a question. How sharp is a pixel? OK then, how soft is a pixel? That's right they aren't either sharp or soft. They just depict a color on our monitor, printer, etc. based upon the data that has been presented. So then, what causes the apperent sharpness or softness of an image? It's nothing more than the contrast between an adjacent range of pixels for the most part (I'm staying away from tech stuff since it will overly complicate a simple concept).

How do they know to apply this so it will be sharp? They don't. 'We' write 'sharpening' routines to modify the data based upon the way we perceive the image as being sharper. Some manufactures build sharpening algorithms into the camera firmware and pre-process the data prior to storing it, and other manufactures elect to process it out side the camera. Canon decided to allow us to elect where, when, and how to apply this sharpening to our images 'outside' of the camera. Thank you Canon for the D30...

[ October 06, 2001: Message edited by: jray ]

[Ron C.]

JRay has it right on the noise reduction aspect of Canon's CMOS image sensor vs CCD.

"Forget all the crap about pixel size, shape, masks, etc. The reason the D30 sensor is so noise free is partly due to the fact they have the gain amplifiers and A/D converters directly attached at each sensor. This not only removes the amount of extranoius noise that can be picked up by PCB (Printed Circuit Board) traces post sensor, and pre A/D converters, but it also increases the signal to noise ratio of the digital signal by virtue of this. This allows lower voltage devices and buffers to be used to transfer data. Once it's digital it takes a lot to screw the data up if it's clean."

Actually, it's almost entirely due to this reason rather than "partly". It's really that simple though and there is no loss in sharpness as someone implied.

Guys, the chief advantage to CMOS is that CMOS lends itself to putting logic to do DSP functions on the imager much more readily than the processes used for CCD. CCD images remain analog until the charges for each pixel are shifted off the device. In the Canon CMOS design each pixel is converted to digital at the pixel as JRAY describes.

This is a huge advantage from an error perspective(which is what noise is).

Someone said this is like the digital vs film debate. That's a completely the wrong analogy. Those are two completely different technologies whereas CMOS vs CCD are essentially the same technology (silicon base image sensing) using somewhat different technics.

I can imaging a scenario where CMOS looses however. This is the BETA vs VHS scenario. 99% of the worlds imager R&D is going into CCD. That makes it cheaper and more rapid advancement towards meeting "consumer" perceptions such as 35mm equivalent sensor size. CMOS may never reach enough economy of scale to beat it.

From my perspective, who cares - as long as I can get as good or better images. The underlying technology is really irrelevant to the user.

Ron C>

[ October 07, 2001: Message edited by: recurry ]

[Pro Photo Store]

Visit the B&H Pro Photo Store for the best Gear Prices and regular specials!

Brilliant Color for any Digital SLR
White Balance Cap
The Pros Agree on ColorRight
What People Are Saying