CCD Vs CMOS

Swordman said:

Ok, could someone explain to me, in simple terms, why nearly all cameras have CMOS sensors instead of CCD?

I know Leica users used to praise CCD sensors for the colours, and from what I remember, CCDs have quite a few advantages (can't remember them, though).

Click to expand...

Hi,

Difference between CMOS and CCD is readout method.

Modern CMOS tends to have less readout noise than CCD.

Best regards

Erik

Mark Scott Abeln said:

Swordman said:

Ok, could someone explain to me, in simple terms, why nearly all cameras have CMOS sensors instead of CCD?

I know Leica users used to praise CCD sensors for the colours, and from what I remember, CCDs have quite a few advantages (can't remember them, though).

Click to expand...

That’s still a fairly common question, despite the shift to CMOS occurring more than a decade ago for most manufacturers.

Silicon sensors of any kind are not sensitive to color. Digital cameras typically detect color via a color filter array placed over the sensor, and then heavy processing is needed to process the recorded signal into plausible colors. If there are any differences in the color output of CCD and CMOS, it must be due to these and not the basic sensor technology.

Standard JPEG processing also seems to have changed, at least with Nikon, leading to less-saturated and flatter looking images out-of-camera. This allows for more flexible editing on the computer, but if you like the richer color of older cameras, you might want to boost saturation and contrast.

Click to expand...

Hi,

CMOS reads voltage of each pixel in place. CCD 'plops' each pixel charge a few thousand times before allowing it to a measuring station.

With CMOS, Canon and Sony used to have different strategies, Sony using massively parallel conversion on sensor, while Canon using 'flash converters' off sensor.

Latest generation Canon sensors probably use the Sony approach.

I now use a Casio FH100 CMOS from about 2010 for high speed video. It has manual shutter speed control down to 25 microseconds. It reduces resolution for high speed video, I assume so that the camera can handle the processing and/or heating issues.

It was well designed for high speed video and does well for athletics in sunlight.

The shutter is 'rolling shutter' so that it reads out one line at a time when it reads out. The line, say, at the top of the frame is processed first and the line at the bottom is processed last. If you look at an object that changes rapidly, it produces 'Jello Effect' distortions, propellers or golf clubs bend, etc. I was interested in distortions for athletic motions and tested the high speed video to see what the Jello Effect distortions for tennis strokes were. They were small and predictable, OK. But for other high speed motions the delay between the top and bottom may produce significant distortions.

For high speed research video cameras, I believe that they all use global shutters. The CCD technology has a global shutter. Many years ago, I had a B&W developmental video camera that was CCD. It had a separate array and each line of the sensor's elemental detector signals were simultaneously moved off the side of the sensor (by voltage steps) onto the storage array for processing. I have lost touch with CCD technology, but that second array might still be the case that allows the global shutter. ? The CCD sensor + storage array might be a factor that makes the cost higher. ? Maybe the second array and readout voltage steps require more power than the CMOS technology ?

One issue with CCDs was that if a bright light source were in the frame, it would saturate the detectors signals. Those electrons would leak into adjacent detectors especially during the readout process when the sensor's elemental signals were moved off the side of the sensor. (or maybe the detectors even continued to record signal during readout?) This leads to streaks in the picture or video that extend to the side of bright light sources. Maybe that problem has been improved. ?

If the storage array is no longer part of CCDs and the saturation streaks from bright lights have been handled, please correct my information.

Chas Tennis

Chas Tennis said:

One issue with CCDs was that if a bright light source were in the frame, it would saturate the detectors signals. Those electrons would leak into adjacent detectors especially during the readout process when the sensor's elemental signals were moved off the side of the sensor. (or maybe the detectors even continued to record signal during readout?) This leads to streaks in the picture or video that extend to the side of bright light sources. Maybe that problem has been improved. ?

Click to expand...

I saw such a problem with the Nikon D200, it was called corduroy banding.

I also saw it in Canon's 20D of the same period, too, which is their kind of CMOS. Though I remember no recall fuss about it there.

Maybe it's just related to reading the pixels out from the edges of the sensor, which Canon did for a long long time with their CMOS and would be related to their notorious banding at high iso/big pushes.

I believe Sony always did it differently and Canon only relatively recently reverse engineered some of their ways, but if I am in the slightest way wrong I can guarantee a tech nerd will enjoy jumping up and down on this post.

Chas Tennis said:

If the storage array is no longer part of CCDs and the saturation streaks from bright lights have been handled, please correct my information.

Chas Tennis

Click to expand...

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

threw the lens said:

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

Click to expand...

Here's a link to a video explaining how it works (it's actually for the A9).

The point is that not only is it fast, it's no longer read sequentially, so no rolling shutter

The difference between CCD and CMOS is the readout method. With CCD the charge is shifted several thousand times to an external 'measuring station'. With CMOS the charge is read in place.

Most modern CMOS sensors do readout in parallel, having thousands of ACDs, normally two for each column, so a 24 MP 24x36 sensor would have 12000 ACDs working in parallel.

The great advantage of CMOS is reduced readout noise. That improves dynamic range a lot.

Sensors are monochromatic. So sensor design does not affect color. It is the CFA in front of the sensor that delivers the color information.

But, early CCDs were often made by Kodak. So, CFA design was probably affected by Kodak's design principles. Another make was DALSA and they also had CFA designs of their own.

There is a third component, sitting in front of the sensor, that contains UV and IR filter and possibly an OLP filter. That will also affect color rendition.

The reason CMOS won over CCD is mostly that it allows for cleaner images with less readout noise. It is also more tolerant of pixeo defects. As CCDs 'pop'charges along columns. a single pixel error on a CCD kills an entire column. With CMOS a failed pixel is just a failed pixel.

Photo diode designs between CCD and CMOS are very similar

Best regards

Erik

Photodog2 said:

CCD is the artists' choice. You can still get used versions of Leica M9, Leica S2, Pentax 645D, K10D; Fuji S5, etc. Not for the higher numbers are always better, must upgrade or die gear love crowd. More limitations and slower workflow meaning less efficient and capable so not for a lot of professionals who are get paid or go hungry. But magic.

Click to expand...

Pretty good list of CCD color titans, Photodog2. I would add the Nikon D1X, D100, D200, and D60 to it. The Kodak DCS Pro 14n also is frequently suggested, but I haven't personally worked with one yet.

fPrime

Erik Kaffehr said:

CMOS reads voltage of each pixel in place. CCD 'plops' each pixel charge a few thousand times before allowing it to a measuring station.

With CMOS, Canon and Sony used to have different strategies, Sony using massively parallel conversion on sensor, while Canon using 'flash converters' off sensor.

Latest generation Canon sensors probably use the Sony approach.

Click to expand...

How does that affect color output? How can that affect color?

Swordman said:

Ok, could someone explain to me, in simple terms, why nearly all cameras have CMOS sensors instead of CCD?

I know Leica users used to praise CCD sensors for the colours, and from what I remember, CCDs have quite a few advantages (can't remember them, though).

Click to expand...

CCD was originally king the only major company using CMOS was canon. Canon worked out the kinks and when the competition saw it they figured out CMOS was best check out some old reviews.

Jonsi said:

Won't give a long, boring, technical explanation of things no one cares about (though others will be glad to), but low light performance is one reason. CCD couldn't match CMOS, and alongside the megapixel race was the ISO race.

I like the CCD results and still shoot mine occasionally.

Click to expand...

They still use CCD for high-end research cameras where CMOS isn't suitable, but the cost of CCD sensors in the 35mm and medium formats are very high.

Mark Scott Abeln said:

Erik Kaffehr said:

CMOS reads voltage of each pixel in place. CCD 'plops' each pixel charge a few thousand times before allowing it to a measuring station.

With CMOS, Canon and Sony used to have different strategies, Sony using massively parallel conversion on sensor, while Canon using 'flash converters' off sensor.

Latest generation Canon sensors probably use the Sony approach.

Click to expand...

How does that affect color output? How can that affect color?

Click to expand...

Yes, you are right.

My posting was not intended to contradict you.

Best regards

Erik

davidedric said:

threw the lens said:

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

Click to expand...

Here's a link to a video explaining how it works (it's actually for the A9).

The point is that not only is it fast, it's no longer read sequentially, so no rolling shutter

Click to expand...

That video does not suggest there is no sequential readout. In fact it says the fast readout "minimizes" rolling shutter which is not "eliminate". The stacked A9 sensor definitely features sequential read out just at a much higher speed than other sensors.

--
Tom

davidedric said:

threw the lens said:

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

Click to expand...

Here's a link to a video explaining how it works (it's actually for the A9).

The point is that not only is it fast, it's no longer read sequentially, so no rolling shutter

Click to expand...

You are mistaken.



The readout is at higher speed. It is still read sequentially. It does not eliminate rolling shutter, though it does reduce it compared to other implementations.

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

As far as I recall Nikon D200 had some kind of electronic shutter that was CCD based.

CCD-s had a small trick which I may think may have allowed for global shutter.

There used to be interline CCDs that had a storage column adjacent to each photosensitive column. So interline CCD could shift the charge to the storage column very fast.

I would think that interline CCDs could reduce rolling shutter. But, the technique means that half the sensor are is used for buffering.

My guess is that most CCDs used for photography were full frame.

https://andor.oxinst.com/learning/view/article/ccd-sensor-architectures

One nice feature of global shutters would be that they would allow for high flash sync speeds.

Best regards

Erik

tbcass said:

davidedric said:

threw the lens said:

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

Click to expand...

Here's a link to a video explaining how it works (it's actually for the A9).

The point is that not only is it fast, it's no longer read sequentially, so no rolling shutter

Click to expand...

That video does not suggest there is no sequential readout. In fact it says the fast readout "minimizes" rolling shutter which is not "eliminate".

Click to expand...

Heck. My response said "pretty much eliminates". You can argue with yourself the difference between that and "minimises".

tbcass said:

The stacked A9 sensor definitely features sequential read out just at a much higher speed than other sensors.

Click to expand...

Again, as I undertand it, there are two processes going on. One is the illumination of the photo sites by the shutter. The second is the transfer of information from the photo sites to the processing layer (and then the transfer of the results to the dram layer for storage).

I don't see why or how the second process, which I believe is the main source of rolling shutter, is sequential.

Though frankly I don't care. I am simply contributing to the discussion, not trying to start a flame war.

Thanks fPrime! I'll keep an eye open on the used market for these CCD classics.

davidedric said:

tbcass said:

davidedric said:

threw the lens said:

davidedric said:

As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

Click to expand...

Do you have a technical reference for that? All I have seen is oscilloscope evidence which suggests the readout is 1/160s, which is unusually fast, they seem to be stacking something in there.

Click to expand...

Here's a link to a video explaining how it works (it's actually for the A9).

The point is that not only is it fast, it's no longer read sequentially, so no rolling shutter

Click to expand...

That video does not suggest there is no sequential readout. In fact it says the fast readout "minimizes" rolling shutter which is not "eliminate".

Click to expand...

Heck. My response said "pretty much eliminates".

Click to expand...

And after that you said 'it does not rely on sequential readout ... it's no longer read sequentially, so no rolling shutter'. Those things are incorrect. It is read sequentially, and there is rolling shutter - just to a lesser degree than in other implementations.

Here's one test that was conducted. There have been others.

'As expected, the A9's electronic shutter was slower [than its mechanical shutter]; we measured the time it took to travel from the top to the bottom of the frame at 6.68 ms, or just under 1/150 second.'

davidedric said:

tbcass said:

That video does not suggest there is no sequential readout. In fact it says the fast readout "minimizes" rolling shutter which is not "eliminate".

Click to expand...

Heck. My response said "pretty much eliminates". You can argue with yourself the difference between that and "minimises".

Click to expand...

You said, and I quote, "As an aside, Sony's stacked CMOS sensor pretty much eliminates rolling shutter, since it does not rely on sequential readout.

davidedric said:

davidedric said:

The stacked A9 sensor definitely features sequential read out just at a much higher speed than other sensors.

Click to expand...

Though frankly I don't care. I am simply contributing to the discussion, not trying to start a flame war.

Click to expand...

Then why keep pushing an incorrect statement? If you're going to contribute try to be factual.

fPrime said:

Photodog2 said:

CCD is the artists' choice. You can still get used versions of Leica M9, Leica S2, Pentax 645D, K10D; Fuji S5, etc. Not for the higher numbers are always better, must upgrade or die gear love crowd. More limitations and slower workflow meaning less efficient and capable so not for a lot of professionals who are get paid or go hungry. But magic.

Click to expand...

Pretty good list of CCD color titans, Photodog2. I would add the Nikon D1X, D100, D200, and D60 to it. The Kodak DCS Pro 14n also is frequently suggested, but I haven't personally worked with one yet.

fPrime

Click to expand...

The Kodak DCS Pro 14n used to be a CMOS camera:

https://www.dpreview.com/articles/3956632068/fillfactorycmos

Best regards

Erik