CCD vs CMOS, again

[hap]

I have a D90 around. Was devastated that it did not have a CCD sensor to play with.

 

[hap]

Here are some photos taken with my Digilux 2 and its CCD sensor. They seem to me to have a lush quality that I don't remember ever seeing from other cameras. I don't know if it's the CCD sensor that made this difference, or the way the Digilux 2 processed the images.

I didn't do any enhancements in post. If any of these photos strike you as, well, striking, at least as far as color goes, could you say what makes them that way? In my mind they stand out from the usual stuff. I believe it's partly due to the brilliant Colorado sunshine, and partly to the vivid colors of the subjects themselves. Do you think there could be something else contributing, e.g. the CCD sensor or the in-camera processing? What words come to mind that describe the special character of these colors?

that's a cool town! Where is it in Colorado?

 

[Rob-F]

that's a cool town! Where is it in Colorado?

Leadville.

 

[Pál_K]

The brightly lit streets and dark cloudy background is what what I find most appealing about those photos.

 

[willie_901]

CMOS sensors used today output a Digital Signal that has undergone some processing before it leaves the chip. (The CMOS sensors used in the Kodak DCS 14MPixel cameras output an analog signal). The Output undergoes much more processing before being written to non-volatile memory.
CCD (non-Hybrid) detectors output an analog signal that is digitized off-chip and is stored with a minimum of processing. The fill-factor of CCD detectors is greater than the FSI CMOS detectors used in the M240 and M10..

The output of my CMOS based Nikon Df is to my liking. But- looking at the NEF file, pain in the butt to write code for. So I only wrote code for the CCD M8, M9, and M Monochrom. A few lines of code to make into an image file. I could look at DCRAW to do the Nikon image, but it's already a processed image before it leaves the camera. The CCD- much closer to raw data.

One other factor between detectors, CMOS or CCD, is the specific chemistry and layout of the chip and the effect on spectral response. Kodak introduced Indium Tin Oxide to improve blue response some 20+ years ago, before some of the other manufacturers. Of late, the Infrared Response has been bred out of newer CMOS detectors. The IR response of the Kodak CCD detectors was much higher than the CMOS detectors used now.

Most cameras estimate and apply a bias offset to in-camera, digital-number data before it is written to the raw file (a simple form of in-camera DSP). Nikon does not. Instead they convert all negative digital number values are set to zero (also DSP).


In-camera digital signal processing has little to do with CCD vs CMOS technology. After the data is digitized, the in-camera data (spatial array of digital numbers) can be modified by DSP before the data is written to the in-camera raw (used for in-camera JPEG rendering) or the raw file written to removable storage media.

The most common of these is noise filtering. In-camera noise filtering DSP is automatically invoked in some cameras at high ISO setting. A double Fourier transform of unrendered, raw-file data is diagnostic for DSP noise filtering.



DSP noise reduction in D300 (CMOS) raw files. The shutter times were 1/15, 1/8, 1/4 1/2, 1, 2, 4, 8, and 15 seconds from left to right, and top to bottom. Noise reduction is evident for shutter times linger than 1/8 sec. A lens cap was on the camera to eliminate signals from light.


DSP noise filtering also affects the noise histogram as this plot of noise vs exposure for the D300 shows.



Now, here's data from a D200 (CCD) raw file.




Raw file, DSP noise filtering is independent of CCD vs CMOS.

However, differences in CMOS vs CCD readout schemes means DSP correlated double sampling - a method used to subtract photo-site reset noise is non-destructive in CMOS.

"An advantage of CMOS sensors is their ability to read the state of the pixels non-destructively -- the state of the sensor (its count of electrons) can be read off without affecting that state. This feature can be used to eliminate the reset noise as follows. After the sensor is reset from the previous exposure, the state of the sensor is read, giving a count of the residual electrons of the reset noise while leaving those electrons in place; then, after the exposure, the sensor is read again, the result being the residual electrons plus those added via photon capture during the exposure. Taking the difference of the two readings gives the photo-electron count of the exposure while subtracting off the reset noise. This before-and-after reading of the sensels is known as correlated double sampling (CDS). On CCD sensors, the electrons in a CCD sensel must be extracted from the sensel and transferred to the edge of the sensor in order to do the readout; CDS can be performed in this per-column readout circuitry rather than separately within each sensel. "

As Brian mentioned there are many other DSP methods that modify raw-file data. Pattern read noise subtraction, PRNU compensation (flat field correction) and dark frame subtraction are a few of many.

A recent example of in-camera DN DSP is Leica M11 pixel-binning. CMOS cameras (M11) must perform the binning process off-chip using DSP. But CCD cameras can bin pixels on chip using analog methods. For pixel binning, CCD has a significant advantage over CMOS. The result is SNR improvement is typically higher for CCD.[link]

 

[hap]

Leadville.

According to family records....I have a distant relative lived in Leadville at the turn of the century and kept a saloon going. POssibly as early as late 1880's.

 

[boojum]

How's that new Pixii working out for you? Are you going to return it and stick with your vintage CCDs? I sure wouldn't want to spend $3000 on a camera and not be able to push it up to 11.

Hey, thanks for asking. The Pixii has taken up residence in my home. It takes good pictures and I will bore you with this one of the Rhodies with the Cooke Amotal. The folks at Pixii have tuned the CMOS APS-C sensor pretty well. I have to A-B it with my M9 to see how that works or better with the M8.2 as the sensors are similar in image size. There are still software glitches but that does not hinder its ability to capture a good image, and push it to 11. ;o) Until you try a Pixii you are somewhat limited in criticism, valid criticism. `There is a principle which is a bar against all information, which cannot fail to keep a man in everlasting ignorance — that principle is contempt prior to investigation.´

https://flic.kr/p/2npiPk1​

Lest you think this image was a fluke here is the folder of a bunch of photos of these Rhodies. I find the Pixii to be a fine camera. Among owners I am not alone. Note that the Pixii captures the detail of the Amotal as well as its softness and luminescence.

https://flic.kr/s/aHBqjzSJm1​

 

[Rob-F]

According to family records....I have a distant relative lived in Leadville at the turn of the century and kept a saloon going. POssibly as early as late 1880's.

Maybe they even knew Baby Doe Tabor! Leadville was her home.

 

[Sonnar Brian]

The CCD used in the M9 and M Monochrom, with the S8612 or BG-55 cover glass has sensitivity into the near UV. The Red dye in the Bayer filter does a poor job with UV. I use a UV filter, preferably an L39 or equivalent to get rid of Purple Fringe on fast lenses.

 

[Rob-F]

The brightly lit streets and dark cloudy background is what what I find most appealing about those photos.

Yes it was a magic day, as far as light is concerned. I could believe that had as much to do with it as the CCD sensor. And yet, that Digilux 2 did put out great color!

 

[boojum]

Yes it was a magic day, as far as light is concerned. I could believe that had as much to do with it as the CCD sensor. And yet, that Digilux 2 did put out great color!

This camera has a reputation for fine color. I wish I had a few less cameras than I do as this one is so tempting.

 

[hap]

Maybe they even knew Baby Doe Tabor! Leadville was her home.

Sounds like someone of ill repute. If true then many "knew" her.

 

[hap]

Sounds like someone of ill repute. If true then many "knew" her.

WEll, I have to retract and apologize. apparently, a great pioneer woman.

 

[luuca]

There is no such thing as CCD vs CMOS.

The photo-diode in the photo sites of CCD and CMOS sensors both convert electromagnetic radiation in to photo electrons. The photo electrons accumulated in each photo-site are stored on-chip as an electric charge. After the shutter closes, the charge is converted into analog DC voltage. The result is a spatial array of DC voltages.

Electrons, electrical charge and DC voltages in CCD and CMOS photo sites are no different than electrical charge and DC voltages in any electric (flashlight) or electronic devices (h-fi amp). They possess no aesthetic characteristics. The electrons, electrical charge and DC voltages alone can not be the source of aesthetic differences in rendered images from CCD and CMOS sensor assemblies.

The M9 rendered image perception you prefer is authentic. But it has nothing to do with CCD vs CMOS.

The difference has everything to do with:

• the sensor cover glass optical properties (including the IR filter film)
• the sensor assembly micro-lens array optical properties
• the sensor color-filter array optical properties (the R, G and B filter films)
• the proprietary demosaicking algorithm(s) used to render in-camera JPEGs and, or post-production JPEGs from DNG files.

ok, this post killed the whole romance of the discussion...

 

[olakiril]

The CCD used in the M9 and M Monochrom, with the S8612 or BG-55 cover glass has sensitivity into the near UV. The Red dye in the Bayer filter does a poor job with UV. I use a UV filter, preferably an L39 or equivalent to get rid of Purple Fringe on fast lenses.

I didn't know this and always thought of UV filters as waste of money. Cheers Brian!

 

[K3N]

https://www.43rumors.com/panasonic-again-displayed-an-organic-sensor-prototype/

Maybe this will add to the discussion

 

[boojum]

Regardless of how long these discussions have gone on they still stir interest because there is a perceived higher image quality to the CCD and the CMOS is always playing catch up. I do not understand the hows or whys of sensors but I am convinced, perhaps wrongly, that the CCD generally returns a better image than the CMOS. I will test the Pixii CMOS against the Leica M8.2 CCD, same lens, same ISO, same f-stop to see the results. I would dearly hope we see some sunshine so I can get images with color on them for a change. When I get it done I will post the results, unnamed as to camera and hopefully I can find a way to scrub the EXIF data. It should be interesting, JPEG's SOOC, in each case, no converted DNG's unless that is thought to be a more honest presentation. If so, the conversion would be GIMP and the results posted on Flickr as the board degrades images past the degradation of shrinking the image to 4608 long side. Come to think of it, converting from DNG to JPEG in GIMP will insure that each image has the same conversion. I'll do that.

Any suggestions welcomed.

 

[Sonnar Brian]

The main advantage of the CCD when compared with Front-Side illuminated CMOS sensors is collecting light coming in at sharp angles. Back Side Illuminated CMOS sensors should work just as well as CCD given a set of microlenses and cover glass designed with this goal in mind. The M11 uses a back-side illuminated CMOS sensor. These days- the biggest difference is you could find the data sheets for CCD detectors as they were sold to the scientific market. Most CMOS detectors for use in cameras do not have much in the way of published material.

 

[Sonnar Brian]

I like a company that loves their products.

http://hamamatsu.magnet.fsu.edu/

Hamamatsu still manufactures CCD detectors for the scientific market. A few years ago I was asked to find a good detector for Visible+near IR (VNIR) applications. Their detector had the best available performance.

https://camera.hamamatsu.com/us/en/p...m_cameras.html

And they like the AIS Micro-Nikkor 55/2.8. Looks like their cameras use C-Mount, with a Nikon F->C adapter and the Micro-Nikkor mounted on it.

 

[Sonnar Brian]

https://www.teledyneimaging.com/en/a...ccd/ccd231-42/

Teledyne discontinued their 24x36 32-MPixel CCD, but seems to still have CCDs for the scientific market. They also make BSI CMOS detectors for the same markets.

https://www.teledyneimaging.com/en/aerospace-and-defense/products/sensors-overview/ccd/ccd42-90/

This CCD has 1/5th the readout noise of the KAF-18500 used in the M9 and M Monochrom. Very low readout noise. 3 e[SUP]–[/SUP]