Any technical explanations as to how or why film sharpness looks different from digital sharpness ??

[faberryman]

What's so "astonishing" that anyone would be interested in what this guy wrote?

About the author: Tim Parkin is the editor of a dedicated landscape photography magazine On Landscape and also the CEO of a drum scanning business.

Still not good enough for you? Do you expect the creator himself to chime in?

There is a vast distance between Tim Parkin on the one hand and "the creator himself" on the other, so I would have been happier if someone with technical expertise somewhere between the two had conducted the test and written the article, which by the way dates from 2014.

The test Tim Parkin performed is riddled with problems. I don't have the interest to go through the article line by line and point them out to you. Given your previous posts, I am confident that you are more than competent to do that for yourself. If you need a head start, read the comments and his nonsensical and evasive responses.

As far as his credentials, his status as editor of his landscape website does not clothe him with any scientific expertise. He also runs a two person scanning business with his wife. The title of CEO seems a bit self-aggrandizing. His photography website has not been updated since 2013.

 

[Steven Lee]

@faberryman You are hard to please, my friend! While nothing is perfect, one would expect a scanner operator, competent enough to have repeat customers, to have a reasonable understanding of the resolution limits of different formats. Whatever problems you see in that test, I am sure are not fatal and did not result in a significant influence on the outcome.

Besides, every optimization (including optimizing for resolution) follows the head+tail distribution: it is relatively straightforward to reap the benefits at the "head". But eventually it approaches the point of diminishing returns. What follows is a long and depressing area of oversized effort relative to miniscule gains. Too many technical debates, I find, get tied up in that area. In my humble opinion the linked test did a great job addressing the practical side of that distribution.

 

[Sirius Glass]

Given the author's credentials I should probably trust his numbers more than my guestimates about this. Since analog resolution has a strong dependency on recorded subject matter contrast, there is no such thing as a pixel value equivalence. These insanely high numbers are commonly thrown out in meaningless comparisons.

A few more things about sharpness and resolution: an F/4 lens turns a tiny point light source into a diffraction pattern of roughly 5 µm. A diffraction pattern diameter of 5 µm corresponds to 100 lp/mm. The lens stopped down to F/8 as used in the test creates twice the diameter. This is physics, so no film or sensor can do anything about this, but digital image "improvents" often try to do it anyway. That's where many people go "my Olympus with this and that lens creates 20MP with pixel sharp results!" and we all know, that this is nonsense.

Oh, and there is DoF, which is abysmally low, if you aim for 5µm resolution. These nice Zeiss formulas we all love to use in Dofmaster are for much larger CoC.

In the end there are only two ways to get huge resolutions:

1. Take the sharpest tele lens you can find wide open, and shoot a very flat wall perfectly perpendicular to your lens axis at moderately close range. Mirror lockup (where applicable) and very sturdy tripod don't need mentioning. Don't get too far away from the subject matter either, or diffusion due to air moisture gets in the way.
2. Take any lens on some digital camera with ridiculously small pixel pitch, shoot any subject matter at any distance and let some smart image processor immerse you in a silly illusion about high res results.

The first method will not give you results actually worth looking at, and the actual resolution you get with the second method will not really depend on the recording medium.

OR do as I do. I use the equipment I have the way I want to get the photographs I want and just skip all the theoretical thoughts and measurements. That can work for film or digital.

 

[Helge]

I found two images on petapixel, which seem to support my claim of "subject matter contrast determines film resolution". In this article they make the umpteenth comparison between analog and digital and (among others) compare image resolution from a Mamiya 7 against the output of a PhaseOne IQ180.

If you look at the comparison between IQ180 and Velvia 50, you see very comparable resolution in the "Nikon" label (high contrast), but more or less mush in the Velvia 50 sections with low contrast.

If you look at the comparison between IQ180 and CMS 20, the low contrast areas look very similar, but the CMS 20 "Nikon" label looks a lot crisper.

This is not some strange miracle, but instead follows directly from the requirement from multiple grains to render a low contrast transition.

Edit: Started writing this earlier today. Can see a few of the points are repeated, but I’ll leave them for continuity.

Tim Parkin who did the test has a drumscanner business and therefor had access to that tech.
Drumscanning with a PMT is good for a lot of things, but optimal resolution is not one of them.
I know Tim would be the first to admit this.

In another version of this article, he has a comparison with a line wedge on different types of film, drum scanned compared to photographed through a microscope.
Here the microscope clearly shows how much resolution is left. Which is rather a lot.

But guess what comes close to a microscope? Good camera scanning!

The two films chosen are notorious for having limited dynamic range. Doesn’t look like he metered for the shadows.

The serial number on the 1.2 50mm lens in the test setup, clearly shows an example of invented detail of the digital camera. The numbers are not legible and look like an alien alphabet. While the film shots are either clear or close to easy to read.

I don’t see how this shows how film “has to build up contrast”?

In the end there are only two ways to get huge resolutions:

1. Take the sharpest tele lens you can find wide open, and shoot a very flat wall perfectly perpendicular to your lens axis at moderately close range. Mirror lockup (where applicable) and very sturdy tripod don't need mentioning. Don't get too far away from the subject matter either, or diffusion due to air moisture gets in the way.
2. Take any lens on some digital camera with ridiculously small pixel pitch, shoot any subject matter at any distance and let some smart image processor immerse you in a silly illusion about high res results.

The first method will not give you results actually worth looking at, and the actual resolution you get with the second method will not really depend on the recording medium.

We don’t disagree, other than that low micro contrast = low resolution is inherent to any optical system. It’s simply how physics work.
Dispersion in the gelatine has equivalents in the CMOS sensor. And is often made entirely too much of.

Bringing up contrast of inherently low contrast subject matter is unnatural and looks unnatural.

Should you want to, you can do the same with film now from a really good scan. Except there is probably more low contrast information to dig into.

 

[DREW WILEY]

Rudeofus, Steven - That linked article was a totally and apples to oranges comparison that might look good on the surface. But dig just a few inches deeper and you'll understand why I call it just "filler". It proves nothing axiomatic whatsoever. Change a single variable with either mode of capture and all bets are off.

And even if the playing field was leveled, they're not even comparing a lens to film performance versus a lens-digital response - no - but an scanner-interpolated film result, with it's own additional set of mismatched variables. Probably not even the angles of perspective were the same. Think about it. The types of lenses are different. In the case of the M7, even the format is different. And the article factoring a single kind of film, which happens to be a particular color chrome product where the dye cloud size is known to differ at a certain point along the scale. Almost nothing can be termed objective. It's more like a mismatched equipment review - what do you want to spend your money on? - a riding lawnmower or a jet waterski?

Back in the day I did numerous review articles of my own, and was paid double just because I didn't do shoot-from-the-hip articles like that one. Makes no difference to me who wrote it. If it had been a college term paper, the professor would have called it a snow job.

 

[DREW WILEY]

Now in terms of how masking affects the present topic. I've been doing film masking for over 40 yrs, and it was done in graphics houses for decades before I knew about it. Masking and sharpening tools in digital programs simulate certain results of that, and are faster and more cost effective, but not necessarily better. Besides, general contrast control, with true film masking you can control the area and density of what I would casually term a translucent halo around detail, which is not lossy because all of the original grain character still shows through (given enough enlargement). But analogously named digital tools, whether unsharp masking or edge enhancement, are basically lossy, and if overdone, are more like an outline drawn around edges.

With film masking you can also enhance microtonality in gradation more subtly. But I won't say much here because the entire subject of masking is more of a tool box of its own, containing many potential tools, rather than a single specific tool. The PDF Ralph linked sorta explains an aspect of it relevant to this particular thread, but seems to be written by someone for which this is not yet a highly developed skill set.

 

[Steven Lee]

@DREW WILEY As I said above, if your goal is squeezing maximum out of your medium and your equipment, online tests like this are of little value. You have to work with your optics, your materials and equipment. But for common print sizes, average skill level, and typical viewing distances such tests give a decent approximation for what to expect.

It is quite common for people with high-end scanners to scan at lower resolutions on purpose. Just to save time, because they are not printing murals. You said elsewhere that HP5+ is too grainy for you in sizes smaller than 8x10. Do you realize how uncommon this is? I mean there's nothing's wrong with your preferences, enjoy what you do. But us mere mortals need to be standing few inches from the wall wearing glasses to be disturbed by HP5 grain on a 24x24" print from a 6x6 neg. And I have enough wall room for maybe two (!) prints of that size in my house.

Why look at everything through a maximalist lens? We're not teenagers anymore

 

[DREW WILEY]

The point isn't my printing preferences versus the "ordinary", but in this case, the fact that all the variables have to be optimized and equally matched for any kind of objective scientific comparison. Otherwise, it's just winging it, BS-style. A low quality scan would simply erase any hypothetical distinction this particular discussion thread is all about. In fact, any scan of the film would end much of the objectivity outright because you're introducing yet another variable beyond those under alleged comparison. ...

Next thing you know, we'll be debating whether a cell phone picture looks sharper or a Velvia film shot run through a really lousy scanner and exported on a low-Res JPEG. Just another waste of time and words.

 

[Steven Lee]

The fact that all the variables have to be optimized and equally matched for any kind of objective scientific comparison. Otherwise, it's just winging it, BS-style.

Stop dreaming right there. You are clearly not a scientist. One cannot have an "objective scientific comparison" because reading and comprehending a result of such comparison would require MS in physics. It would be ridiculous and not welcome because MS/Physics is not a requirement to be a successful photographer. There's a huge distance on the spectrum between "objective & scientific" and "BS-style", and that spectrum is where everyone lives and makes images.

When I worked on computer vision systems used to detect lithography failures for 32nm chip manufacturing, sometimes the limiting factor was soil vibrations. If I were to adopt your style of communication and reasoning, I'd be advising (in a very authoritative tone, nevertheless) for every photographer to carry a 10 tonnes of concrete to erect their tripods on to avoid unnecessary blur. Otherwise their experiences and results should be condemned as amateur BS-style.

When it comes to pixel peeping there's no limit. It helps to have one. The linked article strikes a good balance.

 

[Helge]

Tims article is not scientific nor does it really purport to be.
But it does have the veneer of a science paper that might confuse some people. And perhaps Tim is not quite careful enough in telling us about the unknowns and weaknesses.

We have no idea how careful he was with the framing and matching FoV for example.

It’s not a completely useless piece though.
It has some nice points and shows us for example what 80MP can mean.
And it compares 80MP to a drum scan done by an experienced operator.

 

[grain elevator]

general contrast control, with true film masking you can control the area and density of what I would casually term a translucent halo around detail, which is not lossy because all of the original grain character still shows through (given enough enlargement). But analogously named digital tools, whether unsharp masking or edge enhancement, are basically lossy, and if overdone, are more like an outline drawn around edges.

Just to not leave this unopposed for posterity - this is not true. There's nothing inherently lossy about digital sharpening, any more than darkroom unsharp masking. Both can be overdone, and data is indeed lost if areas are over- or underexpose into featurelessness as a result - in both worlds.

 

[faberryman]

Now in terms of how masking affects the present topic. I've been doing film masking for over 40 yrs...

The usual diversion to masking we have come to expect. If the only tool you have is a hammer, you tend to see every problem as a nail.

 

[Rudeofus]

Tims article is not scientific nor does it really purport to be.
But it does have the veneer of a science paper that might confuse some people. And perhaps Tim is not quite careful enough in telling us about the unknowns and weaknesses.

I didn't link to it to make "that final authoritative statement about film/digital resolution". I already argued, why trying to make such an "authoritative statement" is an exercise in futility. Regardless of how well or poorly done these drum scans are: they do show a profound difference in film resolution between high and low contrast areas, and this difference almost doesn't exist for digital sensors. What you see here is not a scanning artifact, but plain physics, as I have also already explained before. A decent digital sensor cell can hold 10k - 60k electrons, and there is no way that any film grain could hold this many silver atoms and still create measurable density difference between 10.000 and 10010 silver atoms. That's where you need dozens of grains to hold the same information as one sensor cell. If your subject matter is 0% dark or 100% white, then one grain holds the same information as one sensor cell, and sure enough there will be film which easily outresolves any digital sensor ever made. A 1% brightness difference however ...

It’s not a completely useless piece though.
It has some nice points and shows us for example what 80MP can mean.
And it compares 80MP to a drum scan done by an experienced operator.

And it shows yet another time, that once you have a setup, which completely exploits all the resolution possible with either medium, art has died. Film/sensor resolution is an obsession without merit. Both modern film and modern digicams easily outresolve anything artistic you'd ever be able to throw at them. There may be other factors, which tilt the balance more towards film or digital, but resolution certainly isn't, and hasn't been for at least a decade. Please do show me that one image in the world, where people say "wow, I wish TMX was finer grained!".

 

[Rudeofus]

The usual diversion to masking we have come to expect. If the only tool you have is a hammer, you tend to see every problem as a nail.

Masking is way more than one tool. It was used to boost apparent sharpness, but as far as I have read from Drew, he used it much more to clamp down on macro contrast (read: match Velvia 50 with Cibachrome materials).

What I also draw from this discussion: unsharp masking is a very old technique, but basically nobody except for a few very specialized labs would actually use it in the analog era. With digital it has become "the final step in each and every post processing op", and that's why even poorly shot pics started to appear tack sharp. See also: original thread question.

 

[faberryman]

Masking is way more than one tool. It was used to boost apparent sharpness, but as far as I have read from Drew, he used it much more to clamp down on macro contrast (read: match Velvia 50 with Cibachrome materials).

First, don't be so literal. Second, be careful with what you read from Drew. He tends to hyperbolize everything. Not to mention embroider. Third, in every instance, ask a person what he means when he starts throwing around the word "macro contrast". In particular, ask him if he really means "micro contrast".

 

[pentaxuser]

There is a vast distance between Tim Parkin on the one hand and "the creator himself" on the other,

Quite likely but it isn't fair to compare Tim Parkin with Ansel is it?

pentaxuser

 

[Helge]

Regardless of how well or poorly done these drum scans are: they do show a profound difference in film resolution between high and low contrast areas, and this difference almost doesn't exist for digital sensors. What you see here is not a scanning artifact, but plain physics, as I have also already explained before. A decent digital sensor cell can hold 10k - 60k electrons, and there is no way that any film grain could hold this many silver atoms and still create measurable density difference between 10.000 and 10010 silver atoms. That's where you need dozens of grains to hold the same information as one sensor cell. If your subject matter is 0% dark or 100% white, then one grain holds the same information as one sensor cell, and sure enough there will be film which easily outresolves any digital sensor ever made. A 1% brightness difference however ...

I don’t see any profound difference. And what little is maybe there (hard to tell what is what with the different methodologies) is likely again, “corrected” away invisibly in the Phase One.

A significant amount of a CMOS cells capacity is lost in various kinds of noise (thermal, readout, dark current, bias etc.) and much is thrown away in the amplification and quantization.

Silver halide grain is, again, stacked. So i don’t really get your fixation on a single grain?

But a single grain almost always has enormous capacity for absorption of photons, converting to silver development points.

I’d like to see your source on the limits.

No sensor cell would be able to reliably measure the difference between 10.000 and 10.010 photons either.
It get lost in noise. So in that sense you could say “on average”.
Trouble is a sensor cell is much larger than grain.

And it shows yet another time, that once you have a setup, which completely exploits all the resolution possible with either medium, art has died. Film/sensor resolution is an obsession without merit. Both modern film and modern digicams easily outresolve anything artistic you'd ever be able to throw at them. There may be other factors, which tilt the balance more towards film or digital, but resolution certainly isn't, and hasn't been for at least a decade. Please do show me that one image in the world, where people say "wow, I wish TMX was finer grained!".

That’s the same as some of the other platitudes I mentioned earlier, like “you can’t discuss art”.
Yet here we are. It’s clearly something that interests people.

Two reasons for that I think.
It’s easy to talk about. While composition and critique of personal work is hard and a touchy subject.
The second is that it does matter.
Resolution, sharpness and the character of it has a psychological impact.
And it matters at lower resolutions than people normally gives credit for.
A 12MP iPhone photo if printed, starts looking blurry already on A4 format.

People do like to study a print up close or zoom in on a screen. It’s natural if you are interested in a photo.
Resolution has a huge impact. But has become sort of a forbidden fruit, that must not be mentioned too much, even if everyone thinks about it.
Is it the single most important technical aspect? No. But it’s neck on neck with others, like dynamics and tonality.

 

[DREW WILEY]

Rudeofus - you are obviously unfamiliar with much of what color printing involved in the 20th, whether in a darkroom setting or in the commercial printing trade even before Cibachrome. Staggering amounts of masking was routine, workmen were expected to master it, and specialized films were made in even quite large sizes with respect to it. A single color dye transfer print might involved more than fifteen sheets of film before even the sheets of printing matrices were involved. Before that there was carbro.

And no, personally, I did not do it even in Ciba with respect to only gross "macro contrast" (an expression I have never personally used before this moment as a response), but also for sake of inherent color repro corrections and numerous other reasons. But since dye clouds behave a little differently than silver grain, it's a more intuitive to speak about masking on this present thread in terms of black and white printing.

The predictable gadfly Faberryman seems to miss the whole point. I use all kinds of tools in the darkroom. Among them, masking is a whole toolbox unto itself, and not a hammer. Well, yeah, back in Ciba days I actually referred to it as a necessary sledgehammer. But in color neg masking you need a very gentle touch. Dye transfer masking is yet another category, and so is masking for black and white purposes, which in some ways is simpler, but still allows for numerous potential options. But all of that is supplementary to many other darkroom and tools and tricks which do not need to be discussed here. And I mask black and white images only about 5% of the time.

I brought up masking in this context for two specific reasons : 1) to indicate how it can indeed be used to affect edge acutance and overall microtonality (not just overall "macro" contrast issues); 2) to make plain that how many things were done long before their digital mimics, and done quite well. And it appears from the header that this thread is indeed still in "analog" forum territory; so it would seem people would not be in such a rush to convert everything into a scanning frame of mind.

 

[Sirius Glass]

Masking covers many different techniques and each one is either different from others or variations of others. Choosing and properly executing the proper form of masking is an art in itself and hardly a hammer.

 

[chuckroast]

There is a vast distance between Tim Parkin on the one hand and "the creator himself" on the other, so I would have been happier if someone with technical expertise somewhere between the two had conducted the test and written the article, which by the way dates from 2014.

The test Tim Parkin performed is riddled with problems. I don't have the interest to go through the article line by line and point them out to you. Given your previous posts, I am confident that you are more than competent to do that for yourself. If you need a head start, read the comments and his nonsensical and evasive responses.

As far as his credentials, his status as editor of his landscape website does not clothe him with any scientific expertise. He also runs a two person scanning business with his wife. The title of CEO seems a bit self-aggrandizing. His photography website has not been updated since 2013.

Demanding credentials or otherwise dismissing claims based on who said them (rather than what was said) is a debating tactic.

Something isn't right- or wrong based on who said it. Something is right or wrong based on the evidence presented. If you have specific objections and can enumerate the problems with this article, that's one thing. But dismissing the work out of hand because he's just a landscape photographer, his title, or the currency of his website is to the side of the discussion and irrelevant.

By this line of reasoning, most of us here (I suspect) shouldn't be called "photographers" because we are not formally trained in sensiometry, chemistry, and/or art.

Sorry, but this is a pet peeve of mine. In the current culture of hyperventilation, people tend to demand credentials and make arguments from authority when they are unable to either articulate or defend their own views.

And, yes, I'd love to hear a thoughtful explication of the many problems with which this article is "riddled". That might well be a path to learning on my part... for which I acknowledge there is great opportunity

 

[chuckroast]

Masking is way more than one tool. It was used to boost apparent sharpness, but as far as I have read from Drew, he used it much more to clamp down on macro contrast (read: match Velvia 50 with Cibachrome materials).

What I also draw from this discussion: unsharp masking is a very old technique, but basically nobody except for a few very specialized labs would actually use it in the analog era. With digital it has become "the final step in each and every post processing op", and that's why even poorly shot pics started to appear tack sharp. See also: original thread question.

More generally, there are a number of edge enhancement methods born in the analog world, some of which made it to digital.

For example, masking can be use to both enhance edge transitions for an appearance of improved sharpness as well as controlling contrast. As Drew points out, this was a common practice in the transparency days and Ciba in particular. But you can also do that chemically with techniques like semistand, Extreme Minimal Agitation, and SLIMT.

I believe - I could be wrong about this - that the whole digital workflow model is actually rooted in the work that lithographers and printers did to produce multilayer masks and that's why tools like PS have the core feature set they do. Again, I could be dead wrong about this.

 

[faberryman]

Demanding credentials or otherwise dismissing claims based on who said them (rather than what was said) is a debating tactic.

Something isn't right- or wrong based on who said it. Something is right or wrong based on the evidence presented. If you have specific objections and can enumerate the problems with this article, that's one thing. But dismissing the work out of hand because he's just a landscape photographer, his title, or the currency of his website is to the side of the discussion and irrelevant.

I didn't bring up Tim Parkin's credentials. Steven Lee and Rudeofus did. I merely pointed out that the credentials they claimed he had didn't imbue him with any special expertise to conduct the test.

By this line of reasoning, most of us here (I suspect) shouldn't be called "photographers" because we are not formally trained in sensiometry, chemistry, and/or art.

I don't think anyone needs to be trained in sensitometry, chemistry, or art to be a photographer, although training in those subjects might be useful. Your comment is a red herring. Some may call it a "debate tactic" but it is really a logical fallacy.

Sorry, but this is a pet peeve of mine. In the current culture of hyperventilation, people tend to demand credentials and make arguments from authority when they are unable to either articulate or defend their own views.

I didn't hyperventilate, or demand credentials, or make arguments from authority. Nevertheless, I agree with you that it is prudent to be circumspect with what you read on the internet. I certainly don't believe everything I read on Photrio (including from some members in this thread), and hope you don't either.

And, yes, I'd love to hear a thoughtful explication of the many problems with which this article is "riddled". That might well be a path to learning on my part... for which I acknowledge there is great opportunity

As I told Steven Lee, I don't have the interest to go through the exercise of explicating the article for your benefit. You are more than capable of doing so for yourself, For starters, you can read the comments to the article and Tim Parkin's responses, and draw your own conclusions.

 

[Steven Lee]

I would be interested in hearing people's thoughts and feelings towards the relationship between apparent sharpness and grain.

Every image is tack sharp and grain is invisible when your reproduction size is tiny. As you begin to enlarge more and more, those lp/mm numbers from your premium lens begin to pay off. At the same time, grain starts to be more visible. At some point you will say "I cannot enlarge this anymore" because one of the following happens:

1. Not enough resolution: you're not revealing any additional detail by enlarging.
2. Grain becomes too objectionable.

Both are subjective and depend on viewing distance, of course. Personally, I find myself bothered by #2 far sooner than by #1. That's one of the reasons I do not pay much attention to resolution on film. I really like the texture effect added to an image by tight & well defined grain, but it breaks apart after a certain enlargement ratio. A lot of film resolution tests are showing impressive capability of film to resolve fine detail, but those blown up samples look unpleasant to me. The image is being torn apart by grain.

Anyone else thinks this way?

 

[DREW WILEY]

Steven - the error some people make is to inspect grain under high magnification or even a microscope, but not in relation to how a given paper might see it instead, and at a specific level of enlargement. Back in graded paper days, certain easel focusing devices came with a deep blue filter accessory to simulate how the paper saw it. But we can get an idea with VC papers simply via test strips, or by viewing the neg atop a lightbox looking through deep blue and green filters respectively. Staining developers like pyro-based ones have been endlessly discussed (or warred over) in relation to how they factor into all of this.

chuck - it's not accidental that certain terminology occurs in digital applications. Scanning and post-scanning programs were introduced to the printing industry first; and they wisely wanted the transition to be comfortable by using analogously-termed features.

 

[Sirius Glass]

Demanding credentials or otherwise dismissing claims based on who said them (rather than what was said) is a debating tactic.

Something isn't right- or wrong based on who said it. Something is right or wrong based on the evidence presented. If you have specific objections and can enumerate the problems with this article, that's one thing. But dismissing the work out of hand because he's just a landscape photographer, his title, or the currency of his website is to the side of the discussion and irrelevant.

By this line of reasoning, most of us here (I suspect) shouldn't be called "photographers" because we are not formally trained in sensiometry, chemistry, and/or art.

Sorry, but this is a pet peeve of mine. In the current culture of hyperventilation, people tend to demand credentials and make arguments from authority when they are unable to either articulate or defend their own views.

And, yes, I'd love to hear a thoughtful explication of the many problems with which this article is "riddled". That might well be a path to learning on my part... for which I acknowledge there is great opportunity

Pet peeves must be fed, exercised and cared for. Otherwise PETA will get after you. Love your pet peeve.