The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

CMOS Look, CCD Look in MF Sensors

bensonga

Well-known member
Richard,

LOL. As you know, my collection of unused gear isn't useless.....I'm just not putting it to the good use it really deserves!

I hope that someday, perhaps when I retire, I'll be able to make use of all this gear. Regardless of what the future holds (60+ megapixel sensors in phone sized cameras?), I'm sure my CFV-16 will still produce decent images on the day I retire. Long live "fat pixels" (or fat CFAs)!

The 645D is ok too. :D

Gary
 
Just to be clear, I don't know what "fat CFAs" means, either, which is why I asked. An earlier post used that term. So maybe we'll both find out. I think, having done some further research, that the poster probably meant by "fat" purer colors (fat within the color spectrum rather than at the thin ends of that color) but I could have that entirely wrong.
It could be he meant fat pixels, which refers to a low-count, large size sensor. 40MP on a 36x24mm is a much higher density of pixels than 40MP on 33x44mm and in turn higher than 40MP 54x40mm (assuming such a sensor ever gets made). By maximizing the size of the pixels, you get better per-pixel performance; for instance the 645Z has 51MP, but in terms of density, it's equivalent to 30MP if you cropped it down to 36x24, meaning it has larger pixels than the A7R or D810, and in turn better performance.

"Fat CFA" on the other hand, could refer to a high-purity color filter being used in front of the sensor. Each pixel on a sensor captures only light intensity, but to assign each pixel a "color", a filter is used to only allow a narrow spectrum of light to reach each one. In other words, to get a blue pixel, you need to filter out all incoming red and green light, but it may not always be beneficial, if even possible, to have 100% color purity for each pixel.
For starters, besides losing an increasingly large amount of light due to absorption, there actually needs to be enough cross-over between colors in order to render certain types of light a more pleasant color, especially fluorescent which has spikes in it's blue-green spectrum. Your average DLSR is often tuned for shooting under available artificial lighting, because more often than not that's how it'll be used, or so the manufacturer assumes - it's better to be safe and get good-enough looking skin tones in poor lighting, than optimizing for color reproduction and then facing a backlash from people who took photos of people under bad light and got zombies.
I was once trying out an IQ180 at a local photo store, and took a photo of my sister under the store's indoor lighting, needless to say I've never shown her the image.
 

Lobalobo

Member
It could be he meant fat pixels, which refers to a low-count, large size sensor. 40MP on a 36x24mm is a much higher density of pixels than 40MP on 33x44mm and in turn higher than 40MP 54x40mm (assuming such a sensor ever gets made). By maximizing the size of the pixels, you get better per-pixel performance; for instance the 645Z has 51MP, but in terms of density, it's equivalent to 30MP if you cropped it down to 36x24, meaning it has larger pixels than the A7R or D810, and in turn better performance.

"Fat CFA" on the other hand, could refer to a high-purity color filter being used in front of the sensor. Each pixel on a sensor captures only light intensity, but to assign each pixel a "color", a filter is used to only allow a narrow spectrum of light to reach each one. In other words, to get a blue pixel, you need to filter out all incoming red and green light, but it may not always be beneficial, if even possible, to have 100% color purity for each pixel.
For starters, besides losing an increasingly large amount of light due to absorption, there actually needs to be enough cross-over between colors in order to render certain types of light a more pleasant color, especially fluorescent which has spikes in it's blue-green spectrum. Your average DLSR is often tuned for shooting under available artificial lighting, because more often than not that's how it'll be used, or so the manufacturer assumes - it's better to be safe and get good-enough looking skin tones in poor lighting, than optimizing for color reproduction and then facing a backlash from people who took photos of people under bad light and got zombies.
I was once trying out an IQ180 at a local photo store, and took a photo of my sister under the store's indoor lighting, needless to say I've never shown her the image.
Right. As I mention above, I think "fat" CFA refers to purity, as in Doug's article.
 

darr

Well-known member
No offense taken. (And by the way, your Hasselblad 50c images are beautiful, the lighthouse at sunset in particular.)
Thank you for the compliments. The lighthouse picture was made at 800 ASA and could not have been made with my P45 under the conditions as they were. It was dark and the alligators were close by as well as other habitat that live in the NWR, but I had opened the box with the 50c that arrived from Japan an hour before, and I wanted to test its higher ASA performance. The P45 would have taken at least four times or more longer to make the shot, and that was not an option I would have taken.

To answer your question, my workflow has changed. I used to scan (typically Provia 100) on the Epson 1680 with Silverfast to JPEG, working hard to fine-tune the image with that software then making minor adjustments in some rudimentary processor such as PaintShop Pro. I've shifted though away from fine-tuning on the scanner, using VueScan at basic settings, again creating a JPEG (or occasionally a TIFF), then ultimately adjusting in Paintshop, frequently with an assist from Topaz.

Not high tech at all in either case, and I hope to increase my Post Processing skills along with my other skills. Of course, Lightroom or Photoshop or both would be a good start, but shooting film, my approach has been to try and get it right in the camera then replicate the film with a scan.
I am familiar with SilverFast from a few years ago, but not VueScan or Paintshop. (I shot 4x5" for many years as a commercial product/food photographer, and it took me a while to make the digital transition which for me started in 2005.) I understand your approach and getting it right on film would be my approach as well. I grew tired rather quickly of the scanning process, and my last financial contribution into the scanning technique was purchasing a Better Scanning film holder, which only made me feel like I wasted hard-earned money and I even made a post about it on the Large Format Photography Forum at the time. As you can tell from my earlier post, I detest a lot of the financial business of photography as I have seen many students (I am a teacher as well) and photography friends get taken advantage of by unscrupulous sales people. This is not to say they are all bad, but when they are, I tend to campaign against them.

I ask about your digital workflow because a lot of the beauty of digital photography IMO, can be found in post processing. This is not to say one can shoot a really bad exposure and make magic out of it, but all the wet darkroom techniques I used from way back when, is much more streamlined and cleaner in the dry darkroom using digital tools. My raw files be they CCD, CMOS or Foveon are just a baseline I work from. When I used SilverFast for my film files, it was not the same as digging into a first generation sensor file. I understand and appreciate all the engineering knowledge you have thus far read and comprehend, but user experience IMO can present a different set of results that engineering reports might display as off the chart or not so good. I remember back in my earliest film days in photography school class 101 and being taught to bracket my Kodachrome exposures. I found art in the film that was exposed a little too dark or a little too light from what the light meter said; it was magic to my eyes! If I had not learned to deviate from the norm, I would have gotten frustrated with photography and stayed as a working artist in advertising which I was not happy with.

If you do not already have a small digital camera to play with the raw files, I would highly recommend that before buying into MF digital. I am a MF shooter because of my love for the glass, and if I can have the best glass, I want the best camera for that glass and for me that is a technical camera. From there the chain goes to the digital back which I equate to film, and I want my film to have the largest latitude that is available within my budget. The 50c IMO is going to help change how MFD photographers shoot, and how they spend their money! :)

Kind regards,
Darr
 
Last edited:

Lobalobo

Member
4x5 film are thick and quite flat even in the standard Epson holder. I have tested it against the Better Scanning glass "holder" and the difference is minor.
Fair enough, but then what is the advantage of the newer scanner? The Epson 1680 was in its day a high-end flatbed art scanner, and I've been told that I'd be unlikely to see noticeable differences with the V700. When the V800 came out, an Epson salesman told me that the key benefits are the LED, which eliminates the need for warmup, and the glass holders. If you are right, and the glass holders don't matter for 4 x 5, which is flat anyway, then I wonder whether the LED will improve image quality (as I don't care about warmup time).


Silverfast is great if you are scanning color negs. For color pos and B&W, Vuescan is great.
Exactly where I came out. I've given up with negative color film inasmuch as I prefer the high contrast of positive film and lack the patience or skill (or both) to get a color negative scan right. And for positives (or B&W negatives), as you say, Vuescan seems just as good. Of course, this approach leaves me little latitude for exposure error, a particular problem when shooting film, which lacks instant feedback. All the more reason I hope MF digital back prices come down to Pentax 645 price levels, or that I win the lottery.
 

Lobalobo

Member
If you do not already have a small digital camera to play with the raw files, I would highly recommend that before buying into MF digital. I am a MF shooter because of my love for the glass, and if I can have the best glass, I want the best camera for that glass and for me that is a technical camera. From there the chain goes to the digital back which I equate to film, and I want my film to have the largest latitude that is available within my budget. The 50c IMO is going to help change how MFD photographers spend their money! :)
Thanks for the thoughtful comments, which I read beyond those excerpted above. As for the excerpted portion, I shoot RAW almost exclusively on my small sensor cameras--including a travel camera that slips into my shirt pocket--because I prefer to apply my own profiles to conversion. So I'm a kindred spirit, on this point. That said, I confess that I do not fine-tune RAW images before conversion, making individual adjustments from the batch conversion only after an image is in jpeg. I know this is not the best approach and before I were to invest in a MF digital system, I'd learn a better one; in fact doing so, is one of the attractions to me of a MF system and workflow. For small sensor cameras (micro 4/3 or smaller in my case), I'm not sure there would be a payoff in ultimate image quality. (And as you say in another portion of your post, working on a scanned image from my 4x5 film is not the same thing.)
 

richardman

Well-known member
If you head over to the GetDPI Large Format Camera forum, I have a boatload of my photos there. I feel lonely, please add your photos :) For years I have avoided the color negs, opting to shoot B&W and slides, but now a committed color negs user - you get the benefits of HUGE dynamic range (that equals or best most MFDB). As you can see from my samples, I am still tweaking the process, but the results are encouraging.

As for scanners, the 1680 is ancient technology. If you worry about the minutia of CCD vs. CMOS, don't even ask the question and just move to 4990 for low budget, and a V700 for medium budget and the latest and greatest V850 if you have ~$800. If you are asking questions about a $12K-$30K back, it makes no sense to talk about a sub $200 scanner vs. $500+ scanner.

The V850 has some advantage over the V700, but the worst way to get info is from a salesman. Just do some googling read the reviews critically (i.e. is the reviewer reading off specs or are they comparing real samples etc.)
 

Lobalobo

Member
As for scanners, the 1680 is ancient technology. If you worry about the minutia of CCD vs. CMOS, don't even ask the question and just move to 4990 for low budget, and a V700 for medium budget and the latest and greatest V850 if you have ~$800.)
Just to be clear, I never suggested that my own work was at a level of refinement to warrant an analysis of distinctions as subtle as CCD versus CMOS. So it's not quite correct to say I "worry" about such a distinction, only that I noticed the distinction, or thought I did, and was curious about it because I know that others' work is sufficiently refined to trouble over such a distinction.

As for scanner choice, yes I can afford an $800 scanner and will spend the money if I think I will see the difference; you suggest that I will and if that's true I will go forward. (I do get some pleasure creating decent images with older technology, I also shoot and scan photos taken on a 1950s Brownie, but if there is a noticeable difference, then it's foolish for me to wait.) As for posting the images to keep you company, I'll certainly think about it (below is a scanned pinhole I took a few years back), and will surely look at yours.
 

carstenw

Active member
Hello again, sorry for leaving an ambiguous comment, and then not responding for a while. Yes, "fat pixels" mean large sensor combined with low pixel count, so the pixel diameter is relatively large, which gives more subtle gradations.

"Fat CFA" on the other hand is, as in Doug's article, a reference to "thicker" or "denser" colour filters, designed for the best possible colour accuracy, as the expense of cutting some light and thus performing less well at high ISO. This was an older design strategy, which has now fallen out of favour, and I suspect this is the real reason for the "CCD look".

When you make the CFA thinner, you get better high ISO, but less accurate colour, and also more metamerism, i.e. indistinguishable colours.
 

Lobalobo

Member
"Fat CFA" on the other hand is, as in Doug's article, a reference to "thicker" or "denser" colour filters, designed for the best possible colour accuracy, as the expense of cutting some light and thus performing less well at high ISO. This was an older design strategy, which has now fallen out of favour, and I suspect this is the real reason for the "CCD look".
Wonder whether the fat CFA will come back into style, at least as an option on some CMOS sensors, when CCD sensors are no longer being manufactured, inasmuch as at least a minority of photographers prefer the look.
 

ErikKaffehr

Well-known member
Hi,

I wouldn't think so, but you never know. The Sony A7s which is intended for video and low light has relatively fat pixels. The main advantage of fat pixels is that they allow for somewhat higher ISOs, the main disadvantage is that they are very prone to aliasing.

If we look at the P45 that has 6.8 micron pixels it needs to be stopped down to f/16 to by and large eliminate colour aliasing. With 9 micron pixels we would need to stop down to f/22, while with 5.2 microns we would see little aliasing at f/11. Both the new CMOS 50 MP and the IQ-280 are at 5.2 microns.

I would argue that a 5.2 micron image downscaled to say 39 MP will always be superior to a 6.8 micron image yielding 39 MP naturally.

Also, keep in mind that the only cases we look at pixels without resizing are:
  • Looking at an image at actual pixels, known as pixel peeping
  • Printing an image at native printer resolution, ignoring size, and looking at that picture at 25 cm range.

With video, resolution is limited to 2 MP (high definition) or 8 MP (4K), so it may be advantageus to use large pixels with heavy OLP filtering. Even better is to use small pixels, without pixel and line skipping and downscale in image processing, but that takes a lot of CPU-power. But, that is the way to do sharp and unaliased video.

Best regards
Erik




Wonder whether the fat CFA will come back into style, at least as an option on some CMOS sensors, when CCD sensors are no longer being manufactured, inasmuch as at least a minority of photographers prefer the look.
 
Last edited:

Lobalobo

Member
Hi, I wouldn't think so, but you never know. The Sony A7s which is intended for video and low light has relatively fat pixels.
Thanks but my post is about fat "CFAs" not fat pixels. "Fat CFA" is not my term but is explained above in another post as narrow-color definition in the filter, allowing finer grades of color distinctions.
 

ErikKaffehr

Well-known member
Hi,

Sorry, I missed that it was "fat CFA" nio fat pixels.

Best regards
Erik

Thanks but my post is about fat "CFAs" not fat pixels. "Fat CFA" is not my term but is explained above in another post as narrow-color definition in the filter, allowing finer grades of color distinctions.
 

ErikKaffehr

Well-known member
Re: CMOS Look, CCD Look in MF Sensors, profiles matter more?

Hi,

I am doing some checks with different calibration tools, ColorChecker Passport, DNG Profile Editor and QPCard software.

What I can see that the profiles generated by these tools may play a more significant role than the sensors themselves.

Another thing that I have seen quite obviously is that at least when using Lightroom changing exposure will shift colour when using Adobe Standard profiles.

This is coming from "hue tweaks" Adobe does colour. It is described pretty well in this article: ChromaSoft: Visualizing DNG Camera Profiles Part 1 , but my experiments confirm it very well.

I need to look into other raw converters, how they handle this.

I plan to post some notes on this in a few days.

Best regards
Erik
 

Lobalobo

Member
Re: CMOS Look, CCD Look in MF Sensors, profiles matter more?

Hi, I am doing some checks with different calibration tools, ColorChecker Passport, DNG Profile Editor and QPCard software. What I can see that the profiles generated by these tools may play a more significant role than the sensors themselves.
Well that is really interesting and perhaps useful to those who have commented on this thread that they see differences in the CCD versus CMOS look, and like the former. If this is a matter of calibration to whatever look is desired, then the CMOS sensor may well be to everyone's taste (at least if the calibration can be done once for a sensor rather than on an image-by-image basis).

Another thing that I have seen quite obviously is that at least when using Lightroom changing exposure will shift colour when using Adobe Standard profiles.
Although this is out of my league (actually all of this is) others have commented that PhaseOne Sensors need to be converted with Capture to preserve fidelity and allow maximum image quality. Even using small-sensor cameras, I see a difference between conversion with native software and third party software. You are the expert here, not me, but I wonder whether this could explain the color shift with exposure change.
 

ErikKaffehr

Well-known member
Re: CMOS Look, CCD Look in MF Sensors, profiles matter more?

Hi,

I am no expert, just a curious person trying to understand what happens.

The test I just made was very simple. I have two shots of the same IT-8 test target shot 2/3 EV apart. I adjust both raw images to match target grey values.

With Adobe Standard profile I get a difference about DE = 1.7. Removing the hue twists from the Adobe Standard profile using Sandy's tools I get a difference of DE=0.32. Now DE 1.7 is not a large difference, but it makes it quite obvious that exposure differences do affect colour rendition, and this is caused by the way profiles are made.

Need to find out how Capture One behaves on the same samples. Update: Checked Capture One and it shows no (or very small) colour changes with exposure change.

Best regards
Erik




Well that is really interesting and perhaps useful to those who have commented on this thread that they see differences in the CCD versus CMOS look, and like the former. If this is a matter of calibration to whatever look is desired, then the CMOS sensor may well be to everyone's taste (at least if the calibration can be done once for a sensor rather than on an image-by-image basis).



Although this is out of my league (actually all of this is) others have commented that PhaseOne Sensors need to be converted with Capture to preserve fidelity and allow maximum image quality. Even using small-sensor cameras, I see a difference between conversion with native software and third party software. You are the expert here, not me, but I wonder whether this could explain the color shift with exposure change.
 
Last edited:

torger

Active member
I'm a bit rusty here so forgive me if some of the details here is incorrect, but in large it should be; a DNG/DCP profile makes the base color through a matrix (straight linear conversion from the sensor's RGB channels, multiplied with white balance), and then it can have two color adjustment maps, one "huesatdelta" and one "looktable". Both tables have the same format but they are applied differently.

The first is applied before any adjustments (exposure curves etc), and the purpose of this table is to calibrate color, say if the sensor has a tendency towards green in underexposed areas that can be compensated for. The "looktable" is instead applied after any exposure adjustments and intended for a subjective look.

I think most DCP profiles that Adobe provide via Lightroom has only the look table.

Then one needs to be aware of that any curve/contrast change will cause color shifts. Capture One uses simple RGB contrast curves, which means that if a color has say 10%R, 80%G and 50%B and you apply an S curve R will decrease, G will increase and B stay the same, ie you have a saturation increase and a color shift. Adobe has a slightly different varaint of the RGB contrast curve, but for the most part it looks the same.

You can make a contrast curve in the luminance channel, but that will look dull and desturated to most, so the saturation-increasing (and somewhat color-shifting) film curves are intentional. When you fine hand-tune a profile to make pleasing color rendition it's thus important to realize that it will only look the way you intend with the exact same curve as used when making the profile, and yes also the same exposure of the colors.

Therefore I'm very skeptical about changing film curve from the standard (which you can do in Capture One) and still expect the "magic" hand-tuned colors to be there.
 

torger

Active member
When it comes to comparing CCD vs CMOS I think there's some historic stuff involved, ie there's fat pixel CCDs vs antialiased CMOS. Older CCDs are quite noisy and have certain grain or texture which may be visible at 100%, and amplified by the lack of AA filter. Antialiased CMOS can look too smooth and "plastic" in comparison. On a distance or scaled down that should even out though.

Then there's color profiling, with the MF guys making a good job to make skin tones look good in studio conditions, while traditional CMOS has been more about all-around use. I also think that MF users are more likely to use the native raw conversion tool than DSLR users were, ie you compare your DSLR with Lightroom conversion to your Hasseblad with Phocus conversion which is not really fair. On my Canon, Canon native software certainly present colors much better than say Lightroom.

Today with CMOS available in various MF incarnations I think the view of that there is a special CCD look will fade away. There may still be a problem to some that they are "too clean" though regarding texture...
 

Nick Devlin

New member
While I chose to use a camera with the Sony CMOS over the Dalsa CCD, I did see a difference in the image out of camera at base ISO that favoured the CCD. Way long ago, the difference could have been chalked up to the AA filters defiling the files on 35mm cameras. But it's more than that. While i have no empirical basis (or desire) to 'prove' it, my experience tells me that the CCD rendered some colours better natively in a way that is challenging to reproduce with the CMOS, and that the CCD files looked somehow snappier out of camera. But *is very subtle*.

Personally, I have found that judicious use of local area contrast enhancement and sharpening in LR can equalize the 'look' substantially. To a level where it does not matter to me and that other, substantial, advantages of the CMOS chip win out.

Colour is a more interesting question. The most graphic illustration i ever witnesses was the Fuji X-Pro1 vs the M9. The Fuji has gorgeous colour. The Leica Ms have struggled mightily with colour. However, when testing Fuji, I shot a series of images in brightly graphittied alleys, to test the colour response. The Leica rendered colours and subtleties I simply could not replicate with the Fuji - even though the Fuji files were luscious and I would never have wanted for more absent the comparison.

The best i can add to this fractious, and ultimately pointless, debate is that, while some difference is real, CMOS is so much better in more important ways (cost, ISO/DR, Liveview) that this functionally ends the debate.

Also, it must be remembered that for work the ends up in print, the actual differences are almost certain to be lost in press process, as they are on the web. The number of people involved in these debates who are making exhibition prints upwards of 16x20 is much smaller than the number of voices...

None of which is to say that a preference for CCD-based cameras, and their output is not a legitimate reason to chose one. Many superb photographers chose these tools because what they produce simply pleases them more and makes them enjoy the work and streamline their process. And there ain't nothing wrong with that.

Only question is when CCD cameras will become collectors' items :)

- N.
 
Top