It's finally good enough

[Rob C]

Yes, language seems to have been perverted into some strange newspeak dialect in which not even the ordinary superlatives seem to be enough, but need to be enhanced by copious quantities of Red Bull, in a world that eternally shifts between suicidal depression and adrenaline rush.

Blame it on the Valley.

Rob C

[Ray]

Yes, language seems to have been perverted into some strange newspeak dialect in which not even the ordinary superlatives seem to be enough, but need to be enhanced by copious quantities of Red Bull, in a world that eternally shifts between suicidal depression and adrenaline rush.

Sorry! That's not my world. When I say 'cool', I mean the temperature is neither cold nor warm, but somewhere in between. 

[dturina]

Sorry! That's not my world. When I say 'cool', I mean the temperature is neither cold nor warm, but somewhere in between. 

Oh, I speak the jargon, but I always feel silly and that episode of South Park with the Marklar aliens comes to my mind. You know, the ones from Marklar, who refer to all people, places and things as Marklar.

[LKaven]

dturina, I feel you are carrying some misconceptions forth into normative claims.  

You suggest that the 5D matched medium format resolution.  Even the 5DII and the D3x are only really marginally there, just marginally.  You go on to suggest that the only benefit of more resolution is to /print bigger/.   Of course there is an element of enhanced MTF when downsampling from a high resolution capture that preserves an added level of detail comfortably in the upper frequencies of your target sample space.  With a native 12MP sensor, the AA filter and artifacts from bayer demosaicking take away from your 12MP sample space in the upper frequencies.  In fact, that's one of the biggest reasons for using a D800.  There is obvious benefit from its resolution in web-sized images.  Try it yourself.  A lot has happened in 5 years.

[dturina]

You suggest that the 5D matched medium format resolution. 

Yes, and that is exactly true. In fact, I couldn't make 645 scans that are even close, not with my scanner, and I searched the net and found out that the best possible drum scans give a very slight advantage to 645, but only with Velvia. I don't understand why this is even a matter of dispute.
 

[Bart_van_der_Wolf]

Yes, and that is exactly true. In fact, I couldn't make 645 scans that are even close, not with my scanner, and I searched the net and found out that the best possible drum scans give a very slight advantage to 645, but only with Velvia. I don't understand why this is even a matter of dispute.

Hi,

Maybe because others got much better results than you suggest to have gotten? The actually measured MTF curve's Nyquist frequency (106.3 cy/mm) of the 5400 PPI scanner I used for 35mm film when combined with the MTF of film, resolved some 85 cycles/mm, which was still  significantly more than the Nyquist frequency of a 5D (60.67 cy/mm).

It wasn't until the Canon 1Ds Mark II (Nyquist frequency at 69.4 cy/mm) that the difference with film scans (with their lower MTF response at lower spatial frequencies) became small enough (and without graininess, and with a better workflow for the 1Ds2) made me decide to switch to Digital capture. The film scans of medium format film still had a 1.6-1.8x advantage, due to their lower output magnification (but only if the camera kept the film flat, and both lens and technique were top notch) and thus higher resolution and more compact graininess.

Cheers,
Bart

[dturina]

Hi,

Maybe because others got much better results than you suggest to have gotten?

This is not really relevant to me; first, because the results other people got with different equipment don't have much relevance for my own experience and workflow. Second, because I am highly skeptical when someone reports results so far from what I am used to seeing.

The fact is, only the large format seems to hold its ground against 35mm dSLRs. Medium format isn't worse, it's just much harder to get the results. You need to have a Nikon 9000, and then you need to keep the film flat, and recently it's even difficult to get proper processing for E6. So, from what I saw, I'd have to invest in an Imacon or a Coolscan 9000 to really get the kind of sharpness and clarity I'm routinely getting from digital. And the thing is, I don't think the film is better. I am actually getting better landscape colors from digital, I prefer the end result. For portraits, I must admit I sometimes prefer slide film, but not by much, and I only find E100G to really be in the game for me.

As I said, when everything is just perfect, medium format film has a slight edge over my digital equipment, but it's only slight, which means invisible in print, but it is so much easier for me to get good results with digital, I find the additional investment of time and work just unacceptable. I have no problem working hard for the results if they are worth the trouble, but in the case of film, I don't feel they are. For instance, I recently ordered prints made from my 645 Reala negatives, from the only remaining E6 lab in Zagreb. I ordered A3 prints. The results were in every way inferior to the A3 prints I made with 5d. And I mean in every way - colors, detail, tonal gradations, general impression of the image, everything. It's not just a matter of some tiny fraction of resolution, but the tonality of the digital image looked vastly superior. OK, someone will say Reala is crap but I don't really think so, it's actually rather good for landscapes, as C41 films go. I could get better colors from E100G, E100VS, Provia or Velvia, but the dynamic range would be so thin it would be difficult, if not impossible, to shoehorn the scene inside it.

I did at one point feel that film is superior in colors and tonal gradations to digital, but that was when "digital" was an old generation small sensor digicam. With Olympus E-1 I stopped shooting film for all intents and purposes, because it was the same or very close in tonality and for smaller prints, up to A3, the difference in resolution that still favored 35mm film just wasn't significant. But with 5d I don't feel the lack of resolution even in B2 prints, which is the largest that I actually printed. The prints came out very detailed, clean and with perfect colors, and this is the point where I think I finally made my mind about film, even large format. The thing is, I don't feel I'm sacrificing quality by not shooting film. I'm consistently getting the colors I want, and I'm getting results printable in very high quality in sizes as big as I find practical.

So yes, I believe one could shoot a frame of 645 Velvia and expensively scan it on a drum scan and show that it outresolves a 35mm dSLR, but this workflow is not even practical for big studios, and certainly not for me. I have a 4990 Epson and could, if need arises, go out and scan something really important on a drum scanner, if the graphical studios who used them didn't switch to digital in the meantime, but the reality of what I am seeing is much greater difficulty of getting the results that are often not even there.

[LKaven]

I don't see where the claim requires that the film be scanned.  To put it another way, we're not comparing a digital photograph with a digital photograph of a piece of film, we're comparing a digital photograph with a piece of film. 

Where I see the difference is in the high-frequency content of the film, and its native MTF, as compared with a native 12MP capture.  The 5D (or the D3/D700 for that matter) uses an OLPF and a bayer array, and the MTF rolls off very sharply in the high frequencies.  The native 12MP capture does not have a response that extends smoothly out to the Nyquist limit. 

I see the deficiencies of the 12MP capture in the common headshot format, in the detail in hair, skin, and clothing.  I see the benefits of shooting at 36MP even at web size.  It isn't just for larger prints.

The best I can say is try it.

[John R Smith]

I don't see where the claim requires that the film be scanned.  To put it another way, we're not comparing a digital photograph with a digital photograph of a piece of film, we're comparing a digital photograph with a piece of film. 

Absolutely. And this where so many of these film / digital comparisons fall flat on their faces. You should always be comparing prints of the same size, with the film one being made in a wet darkroom on a chemical print from the original negative. Then you are looking, as Luke says, for tonality especially in the upper mids and highlight zones. And if such a comparison is done correctly, by definition there is no way that you can make it over the Internet.

Most folks' recent experience of film is in fact a sort of bastard child, neither truly chemical nor truly digital.

John

[dturina]

I don't see where the claim requires that the film be scanned.  To put it another way, we're not comparing a digital photograph with a digital photograph of a piece of film, we're comparing a digital photograph with a piece of film.  

This is a strange position, because eventually you'll want to do something with that information on film, and when you do, the only place where it makes sense to see film as the end product is projecting a slide on the wall. In all other cases you want either a file or a print, where all the comparisons make sense.

Also, I am yet to see a person who will consistently tell which file originated from film, digital back or a dSLR from a web sized resolution. AFAIR this has in fact been tested and it has been proved that even very skilled photographers can't tell the difference and all the talk about "that special quality which can be seen" don't live up to scrutiny. I like to think I'm very good at things such as colors and fine detail and occasionally I couldn't tell a difference between film and digital, or digiback and 4/3 sensor, from web sized pics.

My problem with film isn't the lack of potential quality, it's the amount of extra work needed to get the results I get with three clicks with digital. Developing E6 is an issue, in especially in anything bigger than 35mm, because the one remaining lab in my town doesn't conform to qlab standards, and you can no longer find good scanners around. It's become a specialty item, at a premium price, and even if I was willing to pay it, I don't see the benefits compared to digital. It's just more difficult, more expensive and I can't do everything myself, meaning I have to depend on manufacturers of film and chemistry, availability of labs and availability of scanners. And I really like doing it all at home - except printing, I do that at a professional studio and have the prints mounted on kapafix and protected by a foil; can't really do it at home.

From where I see it, the differences in resolution between 645, today's m43 cameras and today's top dSLR cameras are increasingly less relevant, something like the difference between iPad 2 and iPad3. Yes it has better resolution screen but it's not like it was a huge problem on the old one, to begin with. Of course I would buy the high res version if they're similarly priced, but if the small incremental difference adds several times the price, bulk and trouble, my decision to upgrade becomes less likely.

[Rob C]

Because one person has personal resource/services problems does not in any way affect the ultimate argument about superiority of which medium.

Hell, I can't even get realistic/good E6 processing on the island where I live anymore - film has to be sent to Barcelona and paid for through the nose. Much of my favourite transparency material has vanished form the face of Earth, but having said that, were I able to have my old 500 'blad stuff back again, I'd use it now in place of anything else I have left in the photo department. Oh yes, I'd also need a good 120 scanner... It isn't really all about digital or film; it's also about the pleasure and certainty associated with the tools.

At least, that's  my take on life these days.

Rob C

[dturina]

Because one person has personal resource/services problems does not in any way affect the ultimate argument about superiority of which medium.

Hell, I can't even get realistic/good E6 processing on the island where I live anymore - film has to be sent to Barcelona and paid for through the nose. Much of my favourite transparency material has vanished form the face of Earth, but having said that, were I able to have my old 500 'blad stuff back again, I'd use it now in place of anything else I have left in the photo department. Oh yes, I'd also need a good 120 scanner... It isn't really all about digital or film; it's also about the pleasure and certainty associated with the tools.

At least, that's  my take on life these days.

Rob C

That's all fine, but the realities of supply and workflow remain; if I can't have a reasonably priced medium format scanner that is critically sharp and can extract the information from the film, it all remains a matter of theoretical, not practical performance. I did shoot film when it was practical, which meant I could get qlab processed E6 in my neighbourhood, and I had a very good 35mm scanner (Minolta Dual IV). But I sold the Minolta years ago, the qlab closed and the remaining options are highly impractical, especially since the digital is advanced enough that the advantages of film remain theoretical.
The fact is, I made more keepers with digital in one year than I ever did with film, so I'm not really ready to lament over its demise. I'm certainly not filmophobic, in fact my fridge door is full of it, but I see frequency of use as a good indicator of my opinion of quality. If I don't use it then it's obvious that I don't see a good reason.

One of the main reasons why I stopped shooting the chromes is deterioration of E6 processing here. The labs don't have enough customers and they attempt to make the process more economical ba stretching the life of the chemicals, and some of the rolls I had processed looked like they were developed in stale piss and bleach. Definitely not the Provia I remember from the old qlab days, I can tell you that.

[LKaven]

Daniel, returning to my most important point.

Compare a digital capture from a native 12MP sensor with a capture from a 36MP sensor downsampled to 12MP.  The downsampled capture will show a higher MTF in the upper frequencies, and show more detail to the naked eye.  You will see this even at web size.

Among the reasons for this are (i) the strong OLPF in a 12MP sensor, along with (ii) bayer demosaicking artifacts, both of which cause the native 12MP sensor to roll off in the upper frequencies, well short of what the 12MP sample space could accommodate.  By contrast, the 36MP capture when judiciously downsampled will exhibit more information in the upper frequencies of the 12MP sample space.  Differences will still be evident even at web size.

[dturina]

Daniel, returning to my most important point.

Compare a digital capture from a native 12MP sensor with a capture from a 36MP sensor downsampled to 12MP.  The downsampled capture will show a higher MTF in the upper frequencies, and show more detail to the naked eye.  You will see this even at web size.

Among the reasons for this are (i) the strong OLPF in a 12MP sensor, along with (ii) bayer demosaicking artifacts, both of which cause the native 12MP sensor to roll off in the upper frequencies, well short of what the 12MP sample space could accommodate.  By contrast, the 36MP capture when judiciously downsampled will exhibit more information in the upper frequencies of the 12MP sample space.  Differences will still be evident even at web size.

This doesn't happen. How can I claim that with such certainty? Because I've done it with 5MP files from Olympus E-1, 12MP files from Canon 5d and 12MP files from Olympus E-PL1. Reduced to 1680x1050 native resolution of my monitor, nobody can tell which is which. My wife can never tell which camera I used unless she knows the picture beforehand. I tested this with several people and they can differentiate between stronger and milder sharpening and that's it. So none of what you're saying holds water.

[LKaven]

This doesn't happen. How can I claim that with such certainty? Because I've done it with 5MP files from Olympus E-1, 12MP files from Canon 5d and 12MP files from Olympus E-PL1. Reduced to 1680x1050 native resolution of my monitor, nobody can tell which is which. My wife can never tell which camera I used unless she knows the picture beforehand. I tested this with several people and they can differentiate between stronger and milder sharpening and that's it. So none of what you're saying holds water.

If you'd like, put up the same image shot with the E1, the 5D, and ... well, the D3x, and the D800?  I'll wager that I can tell the difference right away.  Until the D3x came along, I might have been skeptical, but that changed.  Why not just go rent a D800 and then tell us what you think it's good for or not?  It merits something other than a summary dismissal.

[dturina]

If you'd like, put up the same image shot with the E1, the 5D, and ... well, the D3x, and the D800?  I'll wager that I can tell the difference right away.  Until the D3x came along, I might have been skeptical, but that changed.  Why not just go rent a D800 and then tell us what you think it's good for or not?  It merits something other than a summary dismissal.

Well as far as I can tell your challenge has already been answered several years ago here. If my memory serves me correctly, someone made a slideshow with random photos from digibacks and dSLRs and people could vote which they thought was which. To put it short, it was all random guessing, nobody could tell the difference, so all that stuff about Nyquist and MTF and whatever on smaller images, it doesn't really exist. You can't tell a digiback from a Canon 300d on a web resolution.

[LKaven]

Well as far as I can tell your challenge has already been answered several years ago here. If my memory serves me correctly, someone made a slideshow with random photos from digibacks and dSLRs and people could vote which they thought was which. To put it short, it was all random guessing, nobody could tell the difference, so all that stuff about Nyquist and MTF and whatever on smaller images, it doesn't really exist. You can't tell a digiback from a Canon 300d on a web resolution.

If you have a source, then cite it; otherwise your vague recollections have no authority.  The challenge stands.  We're talking about things that are obvious to anyone who uses a D3X/A900 or D800 after a couple of frames. 

Have you ever used a DSLR with a full-frame sensor with more than 12MP?  Or any camera with more than 12MP?

[BJL]

@LKaven,
Usually, the positive claim of a discernable difference is the one that needs evidence; "no difference" is the null hypothesis to be rejected or corroborated by the data. This is even more relevant when there are other clearly discernable differences in cost, weight, the quality and choice of tools available for composition, and so on.

So can you post or link to evidence of discernable differences in web resolution images?

Two rules:
- No avoidable DOF differences: adjust f-stop in proportion to focal length and linear format size to the extent possible.
- No avoidable blown highlights: these are a common but completely avoidable defect with some images taken with smaller formats.

On DOF: with small enough formats, DOF matching is sometimes impossible, and when so, that is a legitimate difference in look. But this is rarely or never the case between 35mm and MF.

On highlight handling: if one wishes to present dynamic range or handling of high subject brightness range as a difference between formats visible in web-sized images, then comparisons with exposure chosen to hold highlights in all cases should reveal any such differences through differences in shadow handling, tonal gradations and such.

[LKaven]

There was some discussion of this a while back amongst us:

http://www.luminous-landscape.com/forum/index.php?topic=60585.0

There are a variety of positive claims going around, including Daniel's.  But here I'm saying that I've heated water to 100C degrees and observed boiling, whereas Daniel is saying "I've never heated water and I wouldn't, because it could never boil.  This was covered /somewhere/ a long time ago to the best of my recollection."  You might say that whether I've truly observed boiling or not is subject to further confirmation, but one should say that there was honest inquiry.

[BJL]

There was some discussion of this a while back amongst us:
http://www.luminous-landscape.com/forum/index.php?topic=60585.0

I think that two different claims are being confused (at least by me):

1) With Bayer CFA sensors of A and B pixels, A>B, the files that you get by downsampling from the A pixel sensor image to a B pixel file with values for all three colors at each pixel (like JPEG or TIFF) can have higher resolution and greater image quality than those from the B pixel Bayer CFA sensor.

2) Some difference is still visible with both files are down-sampled to a web or screen resolution of C pixels, where C is far less than either A or B.

Claim 1 makes sense to me, for reasons I gave in that discussion you link to. The data in that thread seem to deal with A=24MP, B=12MP, with no data on downsampling further, for example to C=2MP as for comparing at screen resolution.

It is claim 2 that I am skeptical of, and where I would like to see evidence.