The Future of CCD Sensors

[sbernthal]

You are right.

[BJL]

You are right.

OK! Doug Peterson, you have your marching orders.


P. S. I am 99.99% sure that Canon uses the same basic Bayer pattern RGGB color filter arrays as everyone else, but that leaves room for differences in how wide a spectrum of light is accepted by each filter color, trading better color information against more signal and so potentially lower noise levels.

[ErikKaffehr]

Hi,

I would say that this may be a bit of blanket statement. I have seen spectral response curves for very few sensors. It is quite possible that DSLR sensors tend to be more optimised for high ISO, but I have not really seen any evidence for it.

I have only experience with Sony cameras and a P45+ back (that is using a Kodak sensor). What I feel in that case is that white balance is far more important than CFA differences. Using individual DNG profiles both cameras renders colours quite similarly.

Testing with a Color Checker the Sony is almost dead accurate and the P45+ less so. Using an IT8 target the Sony is still more accurate, but Capture One comes close if linear curve is used.

My best friend is shooting Canon. I did not make tests on that camera, but I was quite happy with his colour rendition.

Quite possible that MF colour profiles are better adjusted for skin tone rendition in studio lighting, taking electronic flash taken into account.

Also, I am not really sure that there is a clear consensus on what good skin tones actually are. Skin interpretation seems quite different between different images posted here on LuLa forums.

My guess is that if we eliminate as much variables as possible the results may be pretty close. Example:

• Shoot a grey card for colour balance
• Use a ColorChecker and adjust light gray and black to match reference data
• Use same raw processor
• Generate a profile for the camera in question

And I guess the results may be pretty close. It has been the case for me but I have only access to a few sets of equipment, and my interest is mostly related to stuff I actually use.

Best regards
Erik

Ps. Would anyone be interested a length evaluation of the P45+ vs. Sony is here: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/79-p45-colour-rendition

There is an engineering explanation already: the secret is in the CFA optimisation for colour or for ISO. Almost all dslr CMOS have been optimised to provide great iso, but not the IQ250 which is optimised for colour, even Phase One has said it themselves.

So I would not worry.

Leica has not been using state of the art sensors, so it's not surprising that they have had issues with quality. I would not take their CMOS issues as a sign that there's a specific problem with CMOS on the whole.

[torger]

Canon used cmyg in some older compacts, but nowadays it's rggb in all their models.

[sbernthal]

Testing with a Color Checker the Sony is almost dead accurate and the P45+ less so. Using an IT8 target the Sony is still more accurate, but Capture One comes close if linear curve is used.

This is where the engineer and the photographer part ways:

In engineering you are looking for 100% accuracy or as close to it as possible.
In photography, you are looking for the most pleasing picture.
Case and point, Leaf Aptus which had very inaccurate colors, but still the files were very popular.
I remember shooting a catalog for a designer, where about 50% of the items were in completely the wrong color, no matter how hard I tried to push them in the right direction (it happens with some mid-tones). I was expecting furious reaction from the client, but it never came, and the client continued to come back. With painters it's different though, and they want the right colors to appear in the reproduction. A photographer complained in this forum a few days ago about the "washed out" colors of Credo compared to Aptus, because the Credo colors are much more accurate. I'll still take the Credo as I was feeling quite uncomfortable with the Aptus color interpretation, but the question at hand is not measured color accuracy, but how nice do the end result images look.

[eronald]

I don't see color issues with CMOS - I see texture issues. A friend of mine who is much more knowledgable than me and who has authored a known raw converter says the texture smearing problem appeared with the use of what we now call CDS in CMOS sensors - ie. the last dSLRs before that were the original 1Ds and the Nikon D2x.

On the other hand, if you're a supplier to the fashion industry, I expect they paint in the texture when they make up the ads, so it doesn't really matter how awful the CMOS skin tone is.  Except if you are Cooter and you have pride in your work, I guess, in which case you end up buying up Leica's old stock

Edmund

[ErikKaffehr]

Well,

I don't disagree with you, but one of my observations were that those images could be made pretty close, if work flow is calibrated and matched.

My guess is that the profiles provided with C1 for Phase One backs may be better than those provided for say Nikon cameras. Some posters here even found that profiles for the IQ-250 worked better with Nikon D800 than the D800 profile provided by C1.

That said, I am an engineer with photography as a pastime and quite interested in the engineering part of it…

Best regards
Erik

This is where the engineer and the photographer part ways:

In engineering you are looking for 100% accuracy or as close to it as possible.
In photography, you are looking for the most pleasing picture.
Case and point, Leaf Aptus which had very inaccurate colors, but still the files were very popular.
I remember shooting a catalog for a designer, where about 50% of the items were in completely the wrong color, no matter how hard I tried to push them in the right direction (it happens with some mid-tones). I was expecting furious reaction from the client, but it never came, and the client continued to come back. With painters it's different though, and they want the right colors to appear in the reproduction. A photographer complained in this forum a few days ago about the "washed out" colors of Credo compared to Aptus, because the Credo colors are much more accurate. I'll still take the Credo as I was feeling quite uncomfortable with the Aptus color interpretation, but the question at hand is not measured color accuracy, but how nice do the end result images look.

[torger]

Edmund, can you tell more about what texture issues are? Is it aa filter related or something else?

[torger]

I would say that this may be a bit of blanket statement.

I agree, I'm just repeating what many are saying, have not been able to safely verify it myself. There has been a few other "well-known truths" about MF image quality that has turned out to be marketing b*llshit when really drilling down, like the "16 bit DR" thing. So I can't guarantee that it's true, but I find it reasonable... even if not true concerning color vs ISO, it could be true pleasing vs accurate or some other criteria, what's certain is that different sensors have different CFAs, and sometime even the same sensor type can be equipped with different CFAs for different manufacturers.

[eronald]

Edmund, can you tell more about what texture issues are? Is it aa filter related or something else?

I simply don't know. Skin looks different. It's the plastic wrap effect. The really extremely bad cameras go bad on the cheekbones and under the eyes in outdoors imagery. I don't think it's aa, i think it's some electronic filtering that may be a consequence of the *new* CMOS pixel structure.

Edmund

[Jim Kasson]

So you saying that Canon is using CMYG and Phase is using RGGB?

I was sure everybody were using RGGB.

The principles of the paper apply without regard to the two examples chosen. I, like you, think that all the manufacturers we're talking about in this thread use RGGB (or one of the three other RGB Bayer patterns), but that the transmission spectra of the R, G and B filters could be quite different.

Jim 

[ErikKaffehr]

Hi,

Just to make clear, I also feel it is reasonable, it is just that I don't see any real evidence. There was a discussion on DxO mark comparing a Canon with a Nikon, with the Nikon and in that case the Nikon had much less overlap.

I could also figure that some sensors are intended to optimise rendition of skin tone (melatonin) and some are optimised for rendering chlorophyl. Just a wild guess.

Regarding profiles, I have noticed that the standard studio flash profile and the portrait profile for my P45+ are incredible close, what my memory tells me that maximum DeltaE between the two that I have seen on my IT8 target was about 1.0, below perception really. So very small adjustments on color profiles may matter a lot.

Best regards
Erik

I agree, I'm just repeating what many are saying, have not been able to safely verify it myself. There has been a few other "well-known truths" about MF image quality that has turned out to be marketing b*llshit when really drilling down, like the "16 bit DR" thing. So I can't guarantee that it's true, but I find it reasonable... even if not true concerning color vs ISO, it could be true pleasing vs accurate or some other criteria, what's certain is that different sensors have different CFAs, and sometime even the same sensor type can be equipped with different CFAs for different manufacturers.

[TimothyHyde]

For what it is worth to this discussion, here is a post I made on another board comparing the CCD an CMOS in the Hasselblad offerings. 

I've been asked for some comparisons between the H4D-60 and the H5D-50c.  These first three pairs were shot at ISO 100 (which is base ISO for the H5D-50c; base for the H4D-60 is ISO50).  All were shot  within a minute of each other, and from the same stationary tripod.  These are all straight out of LR5 with no corrections or sharpening. 

With the HC 100
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f11_45th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f11_40th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_15th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_15th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_125th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_125th.jpg

When I first used the new camera I had the impression that the colors were hotter--a little more saturated and vibrant--than my older rig, but this isn't evident when you compare them side by side at ISO100.  The H4D-60 exposes 2/3 of a stop higher than the H5D-50c across all settings, but this is probably due to normal camera variation.  I'll be damned if I can tell any difference in resolution or sharpness between these two cameras.   The major differences are, of course, the crop factor and the ability of the H5D-50c to sing at higher ISOs. 

I'll continue to posts comparisons, next at higher ISO and faces. 

[ErikKaffehr]

Hi,

Thanks for sharing experience and images.

Best regards
Erik

For what it is worth to this discussion, here is a post I made on another board comparing the CCD an CMOS in the Hasselblad offerings. 

I've been asked for some comparisons between the H4D-60 and the H5D-50c.  These first three pairs were shot at ISO 100 (which is base ISO for the H5D-50c; base for the H4D-60 is ISO50).  All were shot  within a minute of each other, and from the same stationary tripod.  These are all straight out of LR5 with no corrections or sharpening. 

With the HC 100
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f11_45th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f11_40th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_15th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_15th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_125th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_125th.jpg

When I first used the new camera I had the impression that the colors were hotter--a little more saturated and vibrant--than my older rig, but this isn't evident when you compare them side by side at ISO100.  The H4D-60 exposes 2/3 of a stop higher than the H5D-50c across all settings, but this is probably due to normal camera variation.  I'll be damned if I can tell any difference in resolution or sharpness between these two cameras.   The major differences are, of course, the crop factor and the ability of the H5D-50c to sing at higher ISOs. 

I'll continue to posts comparisons, next at higher ISO and faces. 

[esox]

I'm very surprosed of the low details precision of the 60mpix. I use P65+ and details are more crispy. Bur I see a difference in the green foliage.

[DandA]

I haven't posted here in quite some time but have actively followed this discussion.  Out of necessity, I have to use the right tool for a given job and that mean most often it's the various pro level CMOS based Nikon DSLR's.  Yet when large files are required for clients or shooting for personal use, I always opt for a CCD based camera.  The scientist in me (formal training) does want to know the theory is behind the differences I see between most CMOS and CCD based cameras but the artist/photographer in me knows what I see and like and what works best for the clients I serve.  They too often prefer the final product derived from the CCD based cameras, even when comparable comparisons from both types of cameras are shown to them. 

The present day CCD cameras I utilize are the Lecia M9 and Pentax 645D.  I've also used on a fair number of occasions the Leica S. All three in many respects have a look that I foten don't find with CMOS based cameras no matter how much I work with and treak the files.  Sometimes I can come close and sometimes not.  Many variables involved.

No doubt there are many advantages to utilization of a CMOS chip and if those advantages are absolutely mandatory to use for a task, then of course I won't hesitate, but again if I have a choice, make mine CCD.

In any case, I have an open mind and most welcome the discussion and opinions of all here, regardless of their personal choices and reasoning.  Thats how one learns and grows.

Dave (D&A)

[eronald]

Timothy,

 Thank you.


Edmund

For what it is worth to this discussion, here is a post I made on another board comparing the CCD an CMOS in the Hasselblad offerings. 

I've been asked for some comparisons between the H4D-60 and the H5D-50c.  These first three pairs were shot at ISO 100 (which is base ISO for the H5D-50c; base for the H4D-60 is ISO50).  All were shot  within a minute of each other, and from the same stationary tripod.  These are all straight out of LR5 with no corrections or sharpening. 

With the HC 100
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f11_45th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f11_40th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_15th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_15th.jpg

With HC 50 II
https://dl.dropboxusercontent.com/u/23138385/h4d60_iso100_f8_125th.jpg
https://dl.dropboxusercontent.com/u/23138385/h5d50c_iso100_f8_125th.jpg

When I first used the new camera I had the impression that the colors were hotter--a little more saturated and vibrant--than my older rig, but this isn't evident when you compare them side by side at ISO100.  The H4D-60 exposes 2/3 of a stop higher than the H5D-50c across all settings, but this is probably due to normal camera variation.  I'll be damned if I can tell any difference in resolution or sharpness between these two cameras.   The major differences are, of course, the crop factor and the ability of the H5D-50c to sing at higher ISOs. 

I'll continue to posts comparisons, next at higher ISO and faces. 

[Gel]

FWIW I tested a Phase P25 back against a 5D3.

Same relative lens setup (primes) and found that whilst the colours were exactly the same the phase was slightly sharper. Probably because there was no AA filter to screw things up.

Now of course, these are older sensors and things move on...but if comparing the look and feel of a CCD vs CMOS sensor there wasn't anything in it.

However shoot a headshot with a 1DX and 85L combo vs a H4D50 and 100mm 2.2 and you will see a marked difference.

My biggest concern though is WHO makes the CMOS. Because as someone with experience of the D800, another Sony sensor, the images lacked any sort of soul to them vs a back with larger pixels like the 1DX.

If someone came out with a CMOS version of the P25 (same 9 micron pixel count) I'd be all over that in a heartbeat and I really think the sensor manufacturing is focused on the wrong thing.
I don't need 50mp for my line of work. Even 22mp is overkill.

[torger]

I don't think low res MF is coming back. First it's hard to sell an MF system with lower resolving power than a smaller format, and second it's not really a good idea. The low res MF backs are moire machines.

I find it quite interesting that many both say that they don't need many megapixels and that they don't want any AA filter. The main reason not to have an AA filter (except from simplifying in manufacturing) is to get more resolving power, ie get more out of the megapixels you got. You do that at the cost of aliasing issues like false colors and moire, which for nice finely patterned suit fabrics can be a real pain.

So if you don't need the resolution for your prints, it's still better for image quality to go higher resolution and put on an AA filter. Images that are a bit soft when you pixel peep can also be enlarged to huge billboard sizes without looking ugly up front, ie you don't see any traces of pixel structures or ugly upsizing fractal effects or whatever.

A problem currently is that pixels are deep, which means that the lower pixel pitch you got the closer to perpendicular the sensor needs its light to avoid color casts and crosstalk, which limits optical wide angle designs to strong retrofocus, a real problem for tech cameras today. The next breakthrough I'm waiting for is that large sensors get access to photodiode-at-the-surface pixel designs that already exists in mobile phone sensors and on some compacts. With that I would not mind grossly outresolving my lenses with the sensor -- no more false colors, no more moire.

However, for MF to keep it's attraction I think their best position to be at now is to strike the balance to have both more megapixels and sharper pixel peep pixels than smaller formats. Many talk about color and look, but I really think most are pixel peepers -- "wow, I get more resolution and pixels are still sharper". So we probably won't see any AA filters in MF cameras any time soon either, although that for real image quality it would be an improvement.

Today I use aliasing as a sort of sharpness check with my digital back - if I don't see aliasing my image isn't sharp enough

[hjulenissen]

If you want larg-ish sensor, low-ish MP count _and_ state-of-the-art sensor tech, then the Sony A7S might be something to investigate?

When I listen to music I don't strive for 44100 Hz native D/A converters using no interpolation filter. Rather, I want the best D/A there is, and this usually means a very high sample rate, combined with noise-shaping/dithering and resampling to give good performance in the audible band. Same with image sensors, I expect that we will at some stage have camera sensors that have "spatial resolution" way in excess of what we need (or what lenses can realistically resolve), or what there is normally photon counts to excite at normal noise level. The I expect this sensor to provide very good image quality at sensible image reprodution sizes. See http://ee.usc.edu/faculty_staff/faculty_directory/fossum.htm

-h