What happened to "CCD is better than CMOS"?!

[voidshatter]

Sure the 5Ds/r has a rather 'low' DR of about 12.4, that could be critical in some moments - but I highly doubt Bill Claffs "calculations" of photographic DR especially when he says the IQ180 has a better DR-performance than the 260. (edit: or was that figure of the 180 just directly taken from DxO? It's not really clear)
 That's just nonsense, they should be at least equal and maybe with a slight advantage for the 260 because of its pixel size (not much but a little).

Besides , his definition and calculation of "photographic DR" is completely arbitrary and has no comparative basis with film photography so I don't think that would be a good argument against my last comment.

Have you done any real world comparison at all? If so, then you should have seen enough difference. I have done many, and if I include the tests done by Doug from DT as well, the conclusion is obvious. There is clearly an advantage of DR for these Sony CMOS sensors when you compare them against Canon or CCD.

[Christoph B.]

I'm talking about real world results.
Not number crunching on a spreadsheet.

Re-read my comments, I'm not denying that the modern CMOS in the 100MP back is technologically more advanced, all I'm saying is that it doesn't really matter all that much. I doubt you'll get more "ooh"s and "aaah"s in your next exhibition because of the new CMOS sensor if you've already been working with an 80 or 60mpx back before and I don't think you'll notice the difference when you're working with it either.

What you will notice is the improvement when it comes to higher ISO settings - but the additional Dynamic Range difference doesn't make that much of a difference in the real world any more, ±13.4 is plenty and I highly doubt you'll see a difference in your photos...

PS: CMOS ISO50 vs CCD ISO200 is not a good comparison...and in the last photo the IQ380 looks worse than my P65+ pushed by 4 stops with standard settings in CaptureOne.

[bjanes]

I'm talking about real world results.
Not number crunching on a spreadsheet.

Re-read my comments, I'm not denying that the modern CMOS in the 100MP back is technologically more advanced, all I'm saying is that it doesn't really matter all that much. I doubt you'll get more "ooh"s and "aaah"s in your next exhibition because of the new CMOS sensor if you've already been working with an 80 or 60mpx back before and I don't think you'll notice the difference when you're working with it either.

Dynamic range is only one parameter of sensor quality. Color response is another and it is affected by the choice of the CFA filters used on the sensor as discussed in this in depth comparison of a Nikon and Canon cameras on DXO. Do you want better color rendering or better low light performance?

Other parameters include optical and electronic cross talk, lens cast color shading, and MTF (modulation transfer function) response. I understand that cros stalk is easier to control with CCD than with CMOS, but front side illumination helps to control cross talk. We don't yet know if the 100MP CMOS is FSI how it stacks up on some of these other parameters. Other considerations include live view and first curtain electronic shutter, available only with CMOS. Combining these parameters into a single overall score is problematic since different users would weight them differently. Measured data are fine, but they must be correlated with field testing to determine if the measured parameters correlate with perceived image quality. Thom Hogan has a good post on this matter.

Bill

[Paul2660]

I'm talking about real world results.
Not number crunching on a spreadsheet.

Re-read my comments, I'm not denying that the modern CMOS in the 100MP back is technologically more advanced, all I'm saying is that it doesn't really matter all that much. I doubt you'll get more "ooh"s and "aaah"s in your next exhibition because of the new CMOS sensor if you've already been working with an 80 or 60mpx back before and I don't think you'll notice the difference when you're working with it either.

What you will notice is the improvement when it comes to higher ISO settings - but the additional Dynamic Range difference doesn't make that much of a difference in the real world any more, ±13.4 is plenty and I highly doubt you'll see a difference in your photos...

PS: CMOS ISO50 vs CCD ISO200 is not a good comparison...and in the last photo the IQ380 looks worse than my P65+ pushed by 4 stops with standard settings in CaptureOne.

Hi Christoph, the reason I believe the 380 is at 200 ISO in the test is because it was a long exposure and the LE and the LE mode on the 380 starts at a base of ISO 200.  I think that was a 60 second exposure in that test if I read it correctly.

Paul C

[Christoph B.]

Dynamic range is only one parameter of sensor quality. Color response is another and it is affected by the choice of the CFA filters used on the sensor as discussed in this in depth comparison of a Nikon and Canon cameras on DXO. Do you want better color rendering or better low light performance?

Sure but what if's already good enough to work with? That's why I posed the question whether anyone here ever encountered problems with the 'limited' DR on the recent CCD backs.

Hi Christoph, the reason I believe the 380 is at 200 ISO in the test is because it was a long exposure and the LE and the LE mode on the 380 starts at a base of ISO 200.  I think that was a 60 second exposure in that test if I read it correctly.

In that case it's fair. Or rather it would be fair if the 100mp back was also set to 200 or if they had compared a 'normal' exposure with another 'normal' exposure at a similar low ISO.
Otherwise you're not showing how much performance the back really has and how much better it really is.

Or maybe then the difference would be 0 and it wouldn't entice people to buy it...who knows?

[voidshatter]

In that case it's fair. Or rather it would be fair if the 100mp back was also set to 200 or if they had compared a 'normal' exposure with another 'normal' exposure at a similar low ISO.
Otherwise you're not showing how much performance the back really has and how much better it really is.

Or maybe then the difference would be 0 and it wouldn't entice people to buy it...who knows?

a) If you looked into the RawDigger screenshot, you could see that the highlight details are about the same for the selected sky region. That means the exposure level are about the same even for different ISO settings.

b) It is not my fault using ISO 200. It is a technology limitation. In Long Exposure Mode the IQ380, the lowest possible ISO setting is 200.

c) If you are not comfortable with the dedicated Long Exposure Mode, you could always ignore the warning and insist shooting in normal mode. However the shadow noise will be destructive.

d) See attached plot of dynamic range comparison based on darkframe noise calculations. In the long exposure territory, using CCD is like using a knife in a gunfight. I have no doubt that a knife can still be a fine weapon for certain situations, but life could be a lot easier with a gun instead.

[voidshatter]

That's why I posed the question whether anyone here ever encountered problems with the 'limited' DR on the recent CCD backs.

I am one "victim" who suffered degraded image quality when I moved from a Nikon D800E to a Phase One IQ260 for long exposure shots of sunrise/sunset. I had a thread at getdpi: http://www.getdpi.com/forum/medium-format-systems-and-digital-backs/52844-backlight-landscape-photography-realized-say-no-silhouette.html

The image below was used as a cover for a photography magazine. It was shot by a Sony CMOS sensor (IQ250). No CCD can do this within a single exposure. If you use CCD then you would have to heavily rely on bracketing, which is something not quite compatible with the long exposure noise reduction countdown of CCD, given that the light condition during sunrise/sunset changes fast.

Sure you could try to take off the ND filter and do a short exposure of the foreground, but then you would run into alignment issues due to the slight change of angle of view caused by the extra layer of glass in front of the lens. Give it a challenge if you don't believe in me (so far no one has ever succeeded this challenge): https://www.dropbox.com/sh/5i9dgi756x4yije/AABKcpNvQov8SFzbZuupv3E1a?dl=0

[Christoph B.]

As I said: Show me two files that were shot with the lowest ISO setting at a normal exposure time, then we'll see if there's any real difference

I know that a longer exposure and higher ISO doesn't work well on CCD chips - but have you ever considered that not everyone takes 60s exposures with ISO 200 or 400 on a daily basis?

If the image quality on ISO 50/100 is pretty much the same then I doubt most of 'us' would consider upgrading to a CMOS back if they never use such a long exposure time and/or a high ISO. Personally I shoot most landscape stuff at ISO50-100 and my longest exposures take between 5-10s and my usual print size is either 50x40cm or 80x60cm / 20x16 and 32x24inches.
Is there a real big discernible benefit for me? I doubt that.
Is there a real big discernible benefit for most users of the current CCD backs? I doubt that too.

Or course there are people who'll need it, people who take very long exposures at night or people who want to shoot portraits outside during the early dawn and the late dusk - whatever the reason there are people who might actually need it. But not everyone because it's not better in every situation. Your last comment highlights exactly that.
If you weren't keen on those super-long exposures a CCD would do perfectly well.

The issue you're describing is less due to the max possible Dynamic Range with the sensor but more due to its bad performance with long exposures.

Also I think I do see some strange issues with colours that are between yellow and green - it seems that CMOS doesn't capture greens as well but had the tendency to favour reds and yellows (see below). I think that's very interesting because here you can find a similar issue with the sensor in the Canon CMOS:
https://www.photigy.com/canon-5d-mark-ii-and-phaseone-p25-does-a-physical-sensor-size-make-a-difference/

Also weak on greens but stronger on yellows and reds, it doesn't look balanced and you get a similar result form the 50mpx and 100mpx CMOS sensor... when you look at the tripod in the left corner it's very strong, almost over-saturated - while the green sheet in the window looks very pale and almost mint-green. The IQ380 delivers a more neutral tripod-colour and a saturated green sheet - also the reflection in the 100% crop is much more on the green side while both CMOS-chips 'see' it as yellow.

Again you might think that the problem in on the side of the IQ380 but it's strange that the same effect appears when you compare an old CCD back to a newer CMOS sensor (like in the link above) and the flower was indeed more green than yellow.

Would it be at all possible to correct that with a colour profile or would that mess with the yellow tones overall?

[voidshatter]

As I said: Show me two files that were shot with the lowest ISO setting at a normal exposure time, then we'll see if there's any real difference

I already did. You could download the files by Doug from DT: https://digitaltransitions.com/massive-still-life-shootout/

[eronald]

I think you guys need to realize that out of 14 stops of file at least at least one needs to be kept free as headroom, one or two are lost to channel balancing, and a couple at the bottom can fall victim in some areas to pattern noise or unmatched amp striping. So a 14 bit sensor really gives you 10 bits in real use, which means two stops safety net to do lifts and curves. Two stops is usually enough to save you from an exposure error during a shoot, but it won't really buy back something like a backlit sunset portrait without flash.

Oh, and I forgot, you lose 1 stop at every stop  ISO, so if you set a Phase CCD back to ISO 200  or 400 under unbalanced light you need to expose really carefully because it is now effectively an 8 or 9 bit sensor, as its *objective* native ISO may be something like 50. The Sony will probably go to 1600 because its native ISO is 200.

I'm sure Doug or Synn will explain that a Phase CCD exposed @ ISO 400 still has 14 bits DR, because it's a 14 bit back.  You can choose to listen to that or do your own experiment with a chart, under incandescent where the channels are mismatched, and watch what happens to your blue channel. It won't get any better at sunset.

In fact from my own experience doing available light, 1 stop latitude is really noticeable, once you are in trouble. Of course one can do noise reduction and other tricks but that's never as good as real DR.  Interestingly there is one $100 solution to get an instant DR upgrade under bad light: use a filter to balance the channels.

There are real advantages to fast CMOS backs. And I think there are probably real differences in the CCD look, but for most users they may not outweigh the speed and DR gain with the new chips. I haven't seen anyone here complaining about the last gen of Sony powered backs.

Edmund

[voidshatter]

Interestingly there is one $100 solution to get an instant DR upgrade under bad light: use a filter to balance the channels.

The skyline is not always a straight line for us to use an ND grad filter. Any foreground cut by the ND grad filter will rely on the DR of the sensor to be recovered. Even in the situation that you have a flat horizon, an ND grad filter will block any cloud higher than the sun (see the underexposed cloud above the sun?), forcing you to rely on DR of the sensor.

[eronald]

Even in the situation that you have a flat horizon, an ND grad filter will block any cloud higher than the sun (see the underexposed cloud above the sun?), forcing you to rely on DR of the sensor.

Yunil,not an ND filter, but a warming or cooling filter. The sensor only has one "ideal" color temperature where the channels are balanced and all the RD is really available.

Edmund

[ErikKaffehr]

Hi,

There is a lot of hype in the industry. Zeiss lenses having high contrast vs. Leica lenses having high resolution. Or the other way around.

When 24x36 sensors having 24 MP CMOS sensors started competing with low end MFD, MFD users told us that MFD/CCD had 6EV advantage in DR. Yes, they even told us that they had 16 bit data files although the format actually used only 14 bits. That was misinformation, or with a clear word a lye.

Next CCD proponents explained that CCDs had superior colour to CMOS, fully ignoring the fact that both CCDs and CMOS are monochrome devices. Colour is added by a colour filter array sitting in front of the sensor. Now, it is quite possible that CCD makers would have better colour filter arrays than CMOS makers.

But, colour is decided much more by processing than by sensor design. It starts with white balance. It is easy to shoot a reference white balance card, but do we do that for every shoot? Next thing is that a WB-card shot doesn't really help, we don't want our sunset shot look as  neutral shot balanced for the warm light.  So, white balance algorithm play a very interesting role.

Next thing is camera profiles. Camera profiles do the interpretation of colour. You are supposed to use your Phase One back with Phase One software, aren't you? Well, that software pretty much implement it's own interpretation of colour.

This is a case of Capture One turning bluish purple into blue: http://echophoto.dnsalias.net/ekr/Articles/OLS_OnColor/SimpleCase/

I don't think there is a simple truth. But I am certainly certain monochrome electronic circuit design has nothing to do with colour rendition.

Now, that Phase One starts converting to CMOS, I don't think they talk so much about the advantages of CCD. It may even seem that they eventually passed 14 EV in dynamic range, so they now have a truly 16 bit data format.

It may take some time to establish the new CMOS backs, but I would be surprised to see a lot of marketing from Phase One on the advantage of CCD when their top of line product is CMOS based.

Now, that said, it is quite possible that medium resolution CCD has an advantage on technical cameras. The present generation CMOS sensors are probably designed for DSLRs, and may not work that well with symmetrical wide angle lens designs.

I would suggest that today's sensors, weather CCD or CMOS do a great job, both architectures are good enough, but I am also pretty sure that there is a change of the guards.

Best regards
Erik

[voidshatter]

Yunil,not an ND filter, but a warming or cooling filter. The sensor only has one "ideal" color temperature where the channels are balanced and all the RD is really available.

Edmund

I understand that a warming/cooling filter could balance the R/G/B/G2 channels, but then again even for a straight skyline you still have to deal with any cloud higher than the sun. Also it requires an LCC shot to be made to really correct the color casts of the color filter.

[Christoph B.]

I already did. You could download the files by Doug from DT: https://digitaltransitions.com/massive-still-life-shootout/

The lowest.

ISO 50 vs ISO 50

Not ISO 50 vs ISO 200.

If it's impossible to do ISO50 for 60s exposures then don't do 60s exposures. Use useful information on equal grounds - or maybe use ISO200 on the CMOS back too if you want a fair comparison.
But i think it's cheating to compare the "old" CCD chip with the ISO200 setting vs the new CMOS at ISO50 and then shove the 'comparison' in everyones face. That isn't very honest.
You can simply take two photos of the same scene with the two backs at ISO50 and use a short exposure of 1/200 or whatever.

And what about the greens in the photo? Wouldn't you agree that those are very different and the CMOS backs exhibit a strong tendency for a brownish-tone?

@ErikKaffher: I use a WB card for nearly every shoot, even for landscapes as a point of reference - and to make a fair comparison I would assume that all backs are set to the same white balance as well - and the rest of the image (blues etc) look fairly similar so I don't think that was the issue.
Maybe that explains why a lot of people see a difference between the color rendition of CCD vs CMOS, it may be that either the red/yellow channel on CCDs are very weak or the greens on CMOS are weak. I doubt that PhaseOne would sell backs that aren't properly calibrated or have a faulty camera profile so I'm assuming what you see is what you get and that's the result.
Sure the RGB circuits are "monochromatic" but that doesn't mean that the capabilities of a CCD and a CMOS is the same when it comes to the strength of the signal capture.

So in my humble opinion there is a clear issue with the greens on the CMOS.

[voidshatter]

The lowest.

ISO 50 vs ISO 50

Not ISO 50 vs ISO 200.

I posted ISO 50 vs ISO 50. What make you think it's ISO 50 vs ISO 200? Are the ISO numbers not clear enough in the screenshot? You could always click on the image to enlarge it.

[Christoph B.]

I posted ISO 50 vs ISO 50. What make you think it's ISO 50 vs ISO 200? Are the ISO numbers not clear enough in the screenshot? You could always click on the image to enlarge it.

Really? So those aren't modified? No brightened shadows beyond reason and use? Would you ever do that to a photo? Really really?

The difference in the highlights is negligible.

And as I already stated - even the files on my P65+ look much better when pushed by 4 stops with the standard basic settings in CaptureOne. That's why I don't think your examples are realistic or useful.

[voidshatter]

Really? So those aren't modified? No brightened shadows beyond reason and use? Would you ever do that to a photo? Really really?

The difference in the highlights is negligible.

And as I already stated - even the files on my P65+ look much better when pushed by 4 stops with the standard basic settings in CaptureOne. That's why I don't think your examples are realistic or useful.

Of course dodge and burn was used on these RAW files for comparison (ISO 50 vs ISO 50).

There are so many examples to have the requirement to push the shadows in post-processing because when we take the shot we have to protect the highlight by underexposing part of the image. I understand that you ignore my examples so I bring you more from others:

Chris has a nice review about the Pentax 645Z (which uses a Sony CMOS sensor): http://chrisgilesphotography.com/blog/pentax-645z-review-pt2-dynamic-range/

Philip also has a nice review here about why we want DR: it is simply not practical to use expensive Broncolor or ARRI LED to illuminate the mountains far away! https://www.storehouse.co/stories/o061f

[Christoph B.]

I can't help it - this just looks wrong. Like a cut-out... I get where you're coming from but I wouldn't do that even if I could, it just doesn't work for me and that's less photography and more 'photo-art' to my eyes. If it works for you - well then good for you!
But I prefer setting up lights and doing as much as possible in-camera.


Modified files without a comprehensive guide as to what was changed are NOT an honest basis for a comparison, especially when you don't provide a before and after photo and the neutral images.

[voidshatter]

I can't help it - this just looks wrong. Like a cut-out... I get where you're coming from but I wouldn't do that even if I could, it just doesn't work for me and that's less photography and more 'photo-art' to my eyes. If it works for you - well then good for you!
But I prefer setting up lights and doing as much as possible in-camera.


Modified files without a comprehensive guide as to what was changed are NOT an honest basis for a comparison, especially when you don't provide a before and after photo and the neutral images.

Many photographers have been trying to utilize bracketing and blending with luminosity masks to counter the high contrast scenes beyond the capability of the camera's limited DR. In Zack's case, it's the Canon sensor. Do you think his results are less photography and more 'photo-art' to your eyes? Or you really simply prefer silhouette?

Nowadays with a Sony CMOS sensor (e.g. the IQ3 100MP) you could achieve this kind of result with a single exposure. Isn't it impressive? How can you achieve this with only flash?