New MF platform from Leica or Sinar makes sense.

[eronald]

I agree. Liveview and electronic shutter is completely impossible with CCD technology, which is the reason why no CCD TV cameras were ever made

Edmund

Hi,

Both CCD and CMOS just give an electronic signal, which in the end is interpreted as a voltage fed to an ADC. So CMOS or CCD has nothing to do with color, the same applies to multishot.

There are two advantages of CMOS over present day CCD devices, the first advantage is lower readout noise, which means that the sensor can go further int darks without producing excessive salt and pepper type noise. The other advantage of CMOS over CCD is continous readout. The latter one is usable for live view.

Modern sensors have very good resolution weather CCD or CMOS, so they need very exact focusing. Technical cameras today have a lot of odd technology for focusing, like extemely long calibrated helicoids that can be used with laser distance meters. Having live view on an MFDB would be very helpful in achieving exact focus with focus peaking and actual pixels view. Live view also would integrate well in tethered shooting situations.

Those are the main reasons I think MF makers will switch to CMOS.

Best regards
Erik

quote author=T.Dascalos link=topic=85784.msg697039#msg697039 date=1389374710]
 I know MS well it is in my daily routine… What I doubt is not the mechanical part (surely a piezoelectric crystal can do the shift with either sensors), but whether a sensor with build-in A/D conversion (like Cmos) are can provide "true colour" information to the (exterior) software, the way it happens with CCD (where A/D conversion happens outside the sensor)… In other words, I don't know if the dedicated program to do MS (flexcolor in my case) takes the info it uses after the transistors or after A/D conversion or from the two combined… Surely, if it can be done with Cmos, it will need a new program that won't be compatible with the ones that the makers already have. This all, is only a suspicion I have which I'm not sure about… If one (or you) can be sure on the matter as to enlighten us, it is more than welcome and will be highly appreciated.

[Ken R]

I agree. Liveview and electronic shutter is completely impossible with CCD technology, which is the reason why no CCD TV cameras were ever made

Edmund

Yeah, but most were 3 CCD designs weren't they? I am no scientist but IIRC a color separation prism absorbs much less light than a bayer filter array which absorbs about 2/3 of the light coming in. I know even consumer three ccd video cameras from way back were pretty good even in low light.

But I know what you are hinting at, high quality live view in Digital Cameras with CCD sensors is possible. More so today with so much advancements in technology.

[ondebanks]

I agree. Liveview and electronic shutter is completely impossible with CCD technology, which is the reason why no CCD TV cameras were ever made

In fairness to Erik, he is right when he is referring to the high-megapixel realm of MFD CCDs. It's just a question of scale: pre-HD TV cameras were what, VGA resolution? That's just 0.3 MP. In the same time as it takes to read out a single 30MP CCD image, you could feed ~100 VGA images through the same single register, single amplifier, single ADC readout bottleneck inherent in the CCD design. Of course some CCDs use dual or quadrant readouts to help speed up matters by 2x or 4x, but that's still a very long way from the "massively parallel" readouts of CMOS.

Ray

[ondebanks]

whether a sensor with build-in A/D conversion (like Cmos) are can provide "true colour" information to the (exterior) software, the way it happens with CCD (where A/D conversion happens outside the sensor)… In other words, I don't know if the dedicated program to do MS (flexcolor in my case) takes the info it uses after the transistors or after A/D conversion or from the two combined… Surely, if it can be done with Cmos, it will need a new program that won't be compatible with the ones that the makers already have. This all, is only a suspicion I have which I'm not sure about… If one (or you) can be sure on the matter as to enlighten us, it is more than welcome and will be highly appreciated.

OK, thanks, I'll try. When doing multishot, a program like Flexcolor receives the same array of digitized pixel counts from the camera as you would find in a finished RAW frame. It's all post-amplifier, post-ADC; the usual 1 integer value per pixel, representing the intensity captured through the usual 1 CFA colour filter per pixel. Therefore it doesn't matter to the software what type of sensor delivered the image. I wouldn't expect multishot usage to require a major program rewrite just because of a change to CMOS.

Ray

[eronald]

Is it true that if I refuse to see you, you don't exist?
The one reproach one could make of these CCDs is that they are not full-frame (eat up some fraction of imaging surface with circuitry).
but 10 frames/s liveview is currently easily doable.

http://www.truesenseimaging.com/products/interline-transfer-ccd

Frankly I don't know whether CCD is better than CMOS; but I do know that the imagery coming out of the japanese dSLRs has never reached the quality of the old MF CCDs when the captures are done in good light. And I belong to the older crowd who think that getting it right in physics, and exporting the data to the customer to deal with, is better than endless software massaging in firmware as the japanese dSLR school seem to practice it, the latest example being the strange encoding of the Raws of the Sony A7R.

Edmund

In fairness to Erik, he is right when he is referring to the high-megapixel realm of MFD CCDs. It's just a question of scale: pre-HD TV cameras were what, VGA resolution? That's just 0.3 MP. In the same time as it takes to read out a single 30MP CCD image, you could feed ~100 VGA images through the same single register, single amplifier, single ADC readout bottleneck inherent in the CCD design. Of course some CCDs use dual or quadrant readouts to help speed up matters by 2x or 4x, but that's still a very long way from the "massively parallel" readouts of CMOS.

Ray

[ondebanks]

If I don't see you, you don't exist.
http://www.truesenseimaging.com/products/interline-transfer-ccd

Edmund

Well interline CCDs give you a nice electronic shutter alright, but they still only have up to 4 quadrant readouts, and thus a slow frame rate at higher pixel counts (the biggest Kodak/Truesense one there, the 29MP, manages 4 frames/sec...not great for live view).

And they've never been used in MFD. Maybe because DR is lower than a full-frame type of CCD; and because microlenses are absolutely essential to route the light away from the "dead" areas dedicated to the line transfers and antiblooming - the microlenses make about a 3x - 10x difference in the quantum efficiency. And we all know that microlenses are not friendly to tech-cam lens usage.

Ray

[ondebanks]

Hmm, you edited and extended your post while I was responding to it.

Frankly I don't know whether CCD is better than CMOS; but I do know that the imagery coming out of the japanese dSLRs has never reached the quality of the old MF CCDs when the captures are done in good light.

I agree. But my assessment is that it's not because of CMOS or CCD per se, but because of AA filters (which are only now starting to go out of fashion), and less spectrally selective CFAs, and sensor size.

And I belong to the older crowd who think that getting it right in physics, and exporting the data to the customer to deal with, is better than endless software massaging in firmware as the japanese dSLR school seem to practice it, the latest example being the strange encoding of the Raws of the Sony A7R.

Count me in that older crowd! I detest when manufacturers mess with the RAWs. What is Sony playing at with the A7R? And Nikon, who used non-optional, non-documented low-pass filtering on long exposures. Canon at least do seem to take the proper view of RAW integrity - they don't even subtract the bias offset - respect! Hopefully others too - Pentax, Fuji, Olympus, Panasonic - I don't know how pure their RAWs are so I can't comment.

Ray

[eronald]

Easy editing is what makes the comp superior to the pen!
IMHO, email should also be editable ...

Of course, I perfectly accept that everything you say is true, and yet to be CONVINCED I would wish to be shown a really good CMOS chip. Just one, please.

Edmund

Hmm, you edited and extended your post while I was responding to it.

I agree. But my assessment is that it's not because of CMOS or CCD per se, but because of AA filters (which are only now starting to go out of fashion), and less spectrally selective CFAs, and sensor size.

Count me in that older crowd! I detest when manufacturers mess with the RAWs. What is Sony playing at with the A7R? And Nikon, who used non-optional, non-documented low-pass filtering on long exposures. Canon at least do seem to take the proper view of RAW integrity - they don't even subtract the bias offset - respect! Hopefully others too - Pentax, Fuji, Olympus, Panasonic - I don't know how pure their RAWs are so I can't comment.

Ray

[eronald]

Well interline CCDs give you a nice electronic shutter alright, but they still only have up to 4 quadrant readouts, and thus a slow frame rate at higher pixel counts (the biggest Kodak/Truesense one there, the 29MP, manages 4 frames/sec...not great for live view).

And they've never been used in MFD. Maybe because DR is lower than a full-frame type of CCD; and because microlenses are absolutely essential to route the light away from the "dead" areas dedicated to the line transfers and antiblooming - the microlenses make about a 3x - 10x difference in the quantum efficiency. And we all know that microlenses are not friendly to tech-cam lens usage.

Ray

I'm not sure there are really technical reasons which would limit the number of readouts, or the DR. Re. the Liveview speed, I'd bet that one can design a circuit that subsamples, does less precise A/D, and speeds up the frame rate by an order of magnitude. As for the microlenses, some trick can probably be found to make them flat on top. Or else the "interline" circuitry could be buried inside the chip. My feeling is this technology has been the victim of  underinvestment.

Edmund

[ErikKaffehr]

Hi,

I have seen an exact description of the Sony coding, they use fewer bits at high data numbers, using something like 1800 values to represent 8000 values. I would say it's OK and probably a means of putting 13 bits of information trough a 12 bit processing pipeline.

http://translate.google.com/translate?hl=sv&sl=ru&tl=en&u=http%3A%2F%2Fblog.lexa.ru%2F2011%2F10%2F28%2Fo_lineinosti_raw_i_ettr.html

The other things they do is data compression on block level.

It is data compression and image manipulation. I guess it may have to do with Sony using the same technology for stills and video.

http://translate.google.com/translate?hl=sv&sl=ru&tl=en&u=http%3A%2F%2Fblog.lexa.ru%2F2012%2F12%2F29%2Fo_sortakh_raw_u_sony.html

The author of the article suggests that this could cause problems on steep gradients, if my interpretation of the google translation is correct.

Best regards
Erik

Hmm, you edited and extended your post while I was responding to it.

I agree. But my assessment is that it's not because of CMOS or CCD per se, but because of AA filters (which are only now starting to go out of fashion), and less spectrally selective CFAs, and sensor size.

Count me in that older crowd! I detest when manufacturers mess with the RAWs. What is Sony playing at with the A7R? And Nikon, who used non-optional, non-documented low-pass filtering on long exposures. Canon at least do seem to take the proper view of RAW integrity - they don't even subtract the bias offset - respect! Hopefully others too - Pentax, Fuji, Olympus, Panasonic - I don't know how pure their RAWs are so I can't comment.

Ray

[ErikKaffehr]

Hi,

May be, but one of the proven benefit of CDDs have been the fill factor. Using interline CCD-s half of that would be given up. CCDs also build on transferring charges from cell to cell, so most charges are popped from cell to cell several thousands of times. I don't think ADC-s can be put on CCDs, and it also seems that available external ADC-s are a bit noisy.

This is said to be the reason that Canons have limited DR, the signal from the sensor is very clean but the ADC is noisy. At high ISO the clean signal from the sensor is amplified before ADC, and the amplified signal is still clean, that is the reason Canons perform so well at higher ISOs.

It may be that CCDs could be developed more, but at this stage it seems that CCDs have no (or little) advantages over CMOS, there is little reason to push an old technology over a newer technology when the newer technology is good enough.

Best regards
Erik

I'm not sure there are really technical reasons which would limit the number of readouts, or the DR. Re. the Liveview speed, I'd bet that one can design a circuit that subsamples, does less precise A/D, and speeds up the frame rate by an order of magnitude. As for the microlenses, some trick can probably be found to make them flat on top. Or else the "interline" circuitry could be buried inside the chip. My feeling is this technology has been the victim of  underinvestment.

Edmund

[ErikKaffehr]

Ray,

Just to say, I have a P45+ and I have a lot of issues with color aliasing, so I don't see the lack of OLP filtering as a benefit. The benefit I see with MFD is sensor size.

Regarding spectral selectivity, I would say you are probably right, even if I am quite a bit enigmatic about the colour rendition of my P45+.

This i a small write up I made on aliasing: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/78-aliasing-and-supersampling-why-small-pixels-are-good

Best regards
Erik

I agree. But my assessment is that it's not because of CMOS or CCD per se, but because of AA filters (which are only now starting to go out of fashion), and less spectrally selective CFAs, and sensor size.

[eronald]

Why do people assume the fill factor is better for CMOS than for CCD?

Edmund

[Theodoros]

Hi,

May be, but one of the proven benefit of CDDs have been the fill factor. Using interline CCD-s half of that would be given up. CCDs also build on transferring charges from cell to cell, so most charges are popped from cell to cell several thousands of times. I don't think ADC-s can be put on CCDs, and it also seems that available external ADC-s are a bit noisy.

This is said to be the reason that Canons have limited DR, the signal from the sensor is very clean but the ADC is noisy. At high ISO the clean signal from the sensor is amplified before ADC, and the amplified signal is still clean, that is the reason Canons perform so well at higher ISOs.

It may be that CCDs could be developed more, but at this stage it seems that CCDs have no (or little) advantages over CMOS, there is little reason to push an old technology over a newer technology when the newer technology is good enough.

Best regards
Erik

I thought that the main reason that MFDBs are not very good on higher ISO, is the absence of microlenses, take your P45+ for instance, it has much more noise than the P30+ because the later has micro lenses which help the photons to be directed to the bottom of the well… The other reason of course is the lower nominal sensitivity. My opinion is that CCDs are underestimated for their capabilities in coping with low light because of how MFDBs are tuned, but OTOH, it is also MO that the makers are correct in their approach…  Let me use another example, the classic Dalsa 33mp sensor is considered to be the nearest "opponent" to your 39mp Kodak Sensor ...right? Now this sensor has a little higher sensitivity than the Kodak sensor, but in return, it is about a stop better at higher ISO, I would't hesitate to use a Sinarback 75LV or a Leaf 75/7 on 400 ISO… If this sensor would have a version with microlenses on its pixels, I believe it would gain another 2 stops, thus being able to achieve completely noiseless prints at 1600 Iso… now, if we take in mind that the sensor is eight years old, it would be reasonable to assume that a latest generation sensor, would be able to achieve more than another, maybe two stops of additional high Iso capability…. add to this the fact that we are talking about no NR appliance in the development on the RAW or of the processor when storing the file and things don't look much different than they do with Cmos FF sensors…  But colour accuracy? Can a Cmos sensor compete with a CCD on that?

[BJL]

I agree. Liveview and electronic shutter is completely impossible with CCD technology, which is the reason why no CCD TV cameras were ever made

As you surely know, the CCD's used in DMF cameras are of the "full frame" type, which does not support video out ... which is why interline transfer CCD and frame transfer CCD were developed for video usage. And looking at the specs on dark noise levels, DR, SNR and such for interline CCD vs full frame CCDs at the sites of Kodak/Truesense and Teledyne/Dalsa, a change to interline CCD would clearly be a big step backward in IQ for DMF. Interline CCD was what Sony (and so also Nikon, Pentax and Konica-Minolta) used before switching to CMOS; does anyone want to suggest that at equal pixel and sensor size the comparison favors interline CCD?

Also, let us avoid comparisons that confound the effects of differences in pixel size, sensor size and CFA design choices with comparisons of the core sensor technology.


P. S. I say "Full frame" not "full frame transfer" because the latter is a neologism coined to avoid the confusion caused when people started using "full frame" in reference to sensor sizes.

[BJL]

... I would wish to be shown a really good CMOS chip. Just one, please.

Can you show me a comparisons between a CCD and a CMOS sensor of the same sensor size and pixel size where the CCD comes out ahead?

[eronald]

Can you show me a comparisons between a CCD and a CMOS sensor of the same sensor size and pixel size where the CCD comes out ahead?

Leica DMR. Leica M8. Leica M9 are all superb at base ISO. Image quality clearly ahead of their competition at the time.
Now, where is this wonderful CMOS chip which can render color as well as the old CCDs?

And yes, I agree that CMOS is better on paper, and should work at least as well in practice. Strangely enough it doesn't, and it is time the geeks accepted what the photographers in this forum all seem to know.

Edmund

[Theodoros]

I guess one can't have everything, CCDs are much better in colour and especially in the way they hold colour in even the deepest of shadows, they are sharper and more contrasty, but I will insist that the CCDs of today have been designed with MF requirements in mind (even M9's one which is a cropped conversion with specially designed microlenses of a larger Kodak one), OTOH , Cmos have "different" approach to DR, which favours overexposing (CCDs favour underexposing) which makes it easier to use from "average shooters". I also insist that there is no proven noise advantage of Cmos, but rather it's a makers decision since MF users, traditionally use their cameras under conditions that don't require LL action shooting.

[BJL]

Now, where is this wonderful CMOS chip which can render color as well as the old CCDs?

Color is a matter of CFA design [Edit: and lens characteristics], not CCD vs CMOS, so your examples of Leica cameras with Kodak sensors seems to be mainly a comment on Kodak's expertise and design decisions in CFA's.

As I said two posts back "let us avoid comparisons that confound the effects of differences in pixel size, sensor size and CFA design choices with comparisons of the core sensor technology."

[ErikKaffehr]

Hi,

Do you have any images showing any CCD advantages on comparable sensors? I have seen a few by Erwin Puts, but not much else. Erwin Puts felt at that time that the M9 was behind the competition.

On the other hand, I just made two A2 prints, one from my Sony Alpha 99 and one from my P45+ and I can essentially not tell them apart. OK, I made it difficult, same subject, same time, same processing using the same tools. OK, sky is a bit different, Sony Alpha 99 a bit magenta the P45+ a bit blue. Some yellowish areas of grass more yellowish on P45+, some strains of grass a tiny bit sharper on the P45+ (?!, Maybe!). Some stones sharper on the Alpha 99? In a 17x24" print those differences are incredibly small!


In general, I may have a feeling that the P45+ has an advantage in sharpness, but when I print comparable images I see little difference.


Anyway, I feel that CMOS offers tremendous advantages, but for me the most important one is live view. I just don't see how anyone can ignore live view for precision work.

I had both CCD and CMOS cameras, Konica Minolta Dimage 7D, Sony Alpha 100 and P45+ on CCD and Sony Alpha 700, 900, 55, 77 and 99 on CMOS. So I actually know what I am talking about. That said, I may be wrong, of course.

Best regards
Erik

Leica DMR. Leica M8. Leica M9 are all superb at base ISO. Image quality clearly ahead of their competition at the time.
Now, where is this wonderful CMOS chip which can render color as well as the old CCDs?

And yes, I agree that CMOS is better on paper, and should work at least as well in practice. Strangely enough it doesn't, and it is time the geeks accepted what the photographers in this forum all seem to know.

Edmund