Film vs. Scanning resolution

[Helge]

Not sure what you intended to state here particularly as it applies to the original topic . . .

Is that a criterion now? I thought we where off on a nice tangent?
I was obviously replying to S138s usual direct or indirect love letter to Epson.

 

[flavio81]

So as a semi-tangent to this article, let's say you want to scan 120 film? Is an Epson V800/850 going to give better results than a V700/750? I'm not trolling, but I need to "prepare" my "Chief Financial Officer" for a "purchase approval" for a scanner in the near future.

The output i've see nfrom those scanners (yes, even the v850), even when using glass carriers, doesn't do justice to 120 film. It smears the image...

Tell your CFO you need a true film scanner.

 

[Les Sarile]

Is that a criterion now? I thought we where off on a nice tangent?
I was obviously replying to S138 usual direct or indirect love letter to Epson.

I see, I was wondering since you quoted me too . . .

 

[Les Sarile]

So, as members of this thread, what can we agree on? Where are the points of disagreement? Why? Is there any non-emotional way to resolve these disagreements?
Phil Burton

Obviously scanner performance at least in in terms of the ability to resolve detail is a known value. I believe some of the contention may be what is "practical in real world use".

For instance, 138S posted a reference that

Epson V700 is able to get all IQ Portra 160 is able to record,

And perhaps on the other end Henning Serger posted at https://www.photrio.com/forum/threads/rollei-rpx-25-grain-and-resolution.115244/page-2#post-1523232

Even the best drum scanners cannot resolve all the details on the film. You have a significant loss.

So is a cheaper scanner that is good enough for "real world results" or do you need better then a drum scanner?
I believe in the end, we all have to decide for ourselves what is practical in terms of cost and time/workflow.

 

[grat]

The output i've see nfrom those scanners (yes, even the v850), even when using glass carriers, doesn't do justice to 120 film. It smears the image...

Even though I have a lowly v800 (I guess the useful discussion is over, may as well descend into irrelevance and await Godwin's arrival), "smear" isn't something I associate with the output. That sounds like poorly focused scanning (the lack of adjustable focus is a serious oversight on these things, personal opinion), or badly processed images.

Can you provide an example? I've seen a number of 120 film scans of very high quality done with an Epson. I've seen few bad scans, except for people trying to prove a point.

I suppose people are unlikely to post their bad scans, but if these scanners were truly as horrible as you and Lachlan claim, no one would use them-- and yet, there are commercial photographers who seem to think they're moderately useful.

Even the test scans I've been doing at 6400 PPI, which is (to my mind) a mostly useless resolution given the preceding discussion, look reasonable (not great, merely reasonable) at the completely pointless 100% size on a 1080p monitor at less than 2 foot viewing distance.

You'll note I haven't posted any of these images-- I'm not entirely happy with my MF efforts so far. Even on shots I was certain were in focus, I'm having difficulty finding the in-focus point. Whether it's equipment or technique, I don't know-- I suspect equipment, as most of them were taken with a TLR (Mamiya C33). Or, perhaps the scans are not quite in focus.

 

[grat]

Obviously scanner performance at least in in terms of the ability to resolve detail is a known value.

You know, I'm not sure it is. Most of the efforts to determine scanner resolution are based around an analog test chart. Even some of the claims as the terrible MTF of the Epson appear to be based on a fairly tortured paper that attempted to use the USAF target in a totally different way than it was designed, to calculate the MTF.

I realize the fundamental purpose of a scanner is take an analog image and turn it into a digital file-- and anytime you convert analog to digital or vice versa, things can get a little squirrely-- but I have difficulty believing that Epson could build a scanner, claim it's capable of 6400 PPI, and not get hit by the mother of all class-action lawsuits because it can barely manage 2400 PPI. If the scanner had come out last year, fine-- but the v700/v750 came out in 2006, at a time when film -> digital was all the rage.

Something doesn't add up. Even the original discussion in this thread, produced a large debate as to what "125 lp/mm" actually meant. Silverfast, using their algorithm, suggests that that translates to 3175 PPI. I believe someone in this thread suggested that actually, 6350 PPI would be required, per Nyquist (while shaky on frequency sampling in general, I had the impression that the 125 lp/mm measurement was already taking Nyquist into account, and therefore 3175 would be sufficient).

Much of this feels like a great deal of hand-waving and armchair engineering-- the sort that suggested that you couldn't breathe in a car going 35 mph, or that bumblebees don't have enough lift to fly-- all great theories, but if empirical evidence contradicts the theory, a new theory is required. I'm not dismissing the information in this thread-- I think there's some really solid information, and I do appreciate all the contributions (and if I offended anyone by trying to steer away from hardware, that was not my intent. The thread on LF forums got pretty ugly at one point).

But the basic fact is that people are getting quality scans from the Epson scanners. They can get better ones from a drum scan (very expensive, very fiddly, and equipment is getting scarce), or from a DSLR (I suspect the quality of the lens is a factor here-- but can be *very* fiddly if you want better resolution than the Epson, and the price is either comparable at the low end, or in the "how high is up" category at the X1 end). In theory, dedicated film scanners such as the PlusTek should deliver better results, but no unit since the discontinued Coolscan has received consistently good reviews-- so it's expensive *and* obsolete.

Ultimately, I think a new type of scanner is needed (well, would be useful, at any rate)-- a "flatbed" scanner that uses a square-ish imager with a quality lens, and creates a stitchable mosaic of whatever size film you put in up to 8x10 (or conceivably larger). I think one could be built for under $500 USD. Certainly for less than the cost of a v850.

 

[Les Sarile]

But the basic fact is that people are getting quality scans from the Epson scanners. They can get better ones from a drum scan (very expensive, very fiddly, and equipment is getting scarce), or from a DSLR (I suspect the quality of the lens is a factor here-- but can be *very* fiddly if you want better resolution than the Epson, and the price is either comparable at the low end, or in the "how high is up" category at the X1 end). In theory, dedicated film scanners such as the PlusTek should deliver better results, but no unit since the discontinued Coolscan has received consistently good reviews-- so it's expensive *and* obsolete.

The Epsons are not exempt from delivering quality scans since quality is not simply more detail. But if the film contains more detail not achieved by the Epsons then obviously better (more expensive) scanners can further resolve it. Most obvious in test charts but can get a lot less obvious in "real world images" depending on how low the system wide quality is.

 

[Les Sarile]

I realize the fundamental purpose of a scanner is take an analog image and turn it into a digital file-- and anytime you convert analog to digital or vice versa, things can get a little squirrely-- but I have difficulty believing that Epson could build a scanner, claim it's capable of 6400 PPI, and not get hit by the mother of all class-action lawsuits because it can barely manage 2400 PPI. If the scanner had come out last year, fine-- but the v700/v750 came out in 2006, at a time when film -> digital was all the rage.

I've seen this sentiment stated many times before . . .
I may have said it once or twice too . . .

 

[138S]

But if the film contains more detail not achieved by the Epsons then obviously better (more expensive) scanners can further resolve it.

I totally agree, but please consider this:

1) As the Epson (focus and flatness nailed) resolves 55lp/mm Hor (48 in Vert) on film, then you have to go well beyond a x14 enlargement to notice the superiority of the more expensive scanner, and at that enlargement the film itself will show important flaws making the image not much suitable for a quality pictorial result.

2) Now imagine we have an ultra sharp CMS 20 negative. IQ is very expensive in effective pixels, you have to increase by a 40% the linear resolving power to notice the least practical improvement, so you need around (2900x1.4) 4000dpi effective to notice a minimal improvement beyond x14. Very diminishing returns.

3) Yes, a curled negative in a V750 shifted 2.5mm from the ideal position will deliver just half of the resolving power, this will make a difference visible from x8 enlargement. It is quite easy to have a 2.5mm miss in a 60mm wide film if not playing attention. Pro Labs deliver quite flat film, but not always with home processing. Those stating the V750 was flawed by x4 had a 4x5" sheet falling 4mm in the center (4mm in 100mm) and they were not aware, crazy...


From these points, what makes mostly the difference is the human factor, there are many ways to destroy native IQ in the edition, many home users play repetitive destructive operations in Ps, don't use the required oversampling, they save many times in destructive jpg, no 16bits usage when it's necessary, wrong algorithm setings in the resize, infame sharpening, infame color edition. A Pro service usually has all that quite optimized, and the Pro operator may sport a powerful aesthetic criterion to craft a wonderful image, he is all day long doing that.

Of course a Pro scanning operator won't use much an Epson, a Pro machine is a total quality reference that saves manpower. For top quality, with the Epson an accurate sharpening optimization is required from big files, a Pro machine delivers an optimized image containing just the necessary pixels.


But there is something unfair. Some commercial services show forged/unfair tests to discredit the Epson by posting Epson images from a clumsy customer next to his Pro job. Hey, this is not the truth... Many proficient amateurs may match the Pro service quality with the Epson.

This is the case of Alan Klein, he posted the example, isn't it? It is LF which stresses less film capability, but it was TMX on tripod shot with one of the finest lenses available, an APO Sinorar-S, at best aperture.

Is that a criterion now? I thought we where off on a nice tangent?
I was obviously replying to S138 usual direct or indirect love letter to Epson.

What I love is our pair of 138 tanks, One of them permanently pointing to the wall, and the Drytac press, which is a totally pro gear we don't deserve to own. Finishing a darkroom and preparing for the post covid, we'll place the 8x10" conversion on the 138 pointing to the wall and I guess we'll learn what is a big fine print. Of course size and quality are not related, but a fine 1.4m print from 8x10"...

https://www.photrio.com/forum/threads/darkroom-portraits-part-2.89555/page-16

claim it's capable of 6400 PPI, and not get hit by the mother of all class-action lawsuits because it can barely manage 2400 PPI.

It samples 6400dpi, yielding 2900 effective in the H and 2300 in the V, all flatbeds, or even a flextight has lower performance in the motion direction. All scanners have a loss from sampling and effective, the Nikon is rated effective at the same than the sampling, but this is from flawed concepts, Shannon etc explains the reason.

This happens with all digital gear, smartphones say 40MPix and real yield is 1/4 effective.

Ultimately, I think a new type of scanner is needed (well, would be useful, at any rate)-- a "flatbed" scanner that uses a square-ish imager with a quality lens, and creates a stitchable mosaic of whatever size film you put in up to 8x10 (or conceivably larger). I think one could be built for under $500 USD. Certainly for less than the cost of a v850.

This is a mix between a Frontier (area sensor) and a Creo (automatic stitch). It would be very expensive and perhaps complex to service, but not impossible to make it cheap.

There is another possibility. The V850 has two lenses of ultra high quality working at optimal magnification for the design, the (focal) longuer one delivers 17000 pixels effective in the 5.9" scan width, problem is that it takes 3 or 4 35mm strips at the same time and not one like the hassie.

Let's imagine we replace the V850 lens with another one covering only 3" instead 5.9", we would simply double the performance to 5800 dpi effective which is way beyond we may need for 35mm and 120 film. Of course good holders would be required to nail precise focus.

The Epson V700/+ obtains a suprising yield (17000 effective pixels in the scan width) from using lenses optimized for its fixed working distance, making a lens that's acceptable for a wider range it is more expensive as a battery of optical corrections have to be implemented in the design, the hassie uses a derivative of an enlarger lens that is more expensive (retail) than the whole V700, but it zooms in/out...

The V700 is a very sound design with an incredible yield for the cost, only a powerful corporation like Epson is able of that, it is a very well balanced design, renouncing to zoon to exploit a cheaper lens at its optimal magnification, covering 5.9" to not require more precision in the focus, and still capturing well most ot what film usually contains. Prosumer product concept: mostly pro results with an additional effort from user.

But the V700 design may go further, today people don't buy the Epson to scan opaques... A roll film version can be made by stripping the wide lens, reducing coverage of the HiRes lens to 3" (still scanning 2 35mm strips) and providing a quality holder that allows those 5800 effective dpi, technically blowing away the Plustek and old Nikons, but providing no better practical yield because limits are in the film anyway.

A simpler approach would be simply replacing the 5.9" coverage lens by a 3" or 4" coverage one, with a lower drive ratio and better holders.

 

[SCHWARZZEIT]

I did it...

But to conserve the 2600 effective you cannot downsize it to 2600, you have to add a safety margin related to rayleight/nyquist/shannon/etc criterion.

If you take the original crop and you make a super pixel peeping (of the clock) in Ps you will find that an edge takes around some 4 pixels to make the transition, this suggests that the actual effective performance is less than 1/2 of the scanned resolution, so 2600 or less, by no means it is 4000.

Downsizing has nothing to do with the Rayleigh criterion. While the Shannon-Nyquist sampling theorem also applies to downsizing images, it does not necessarily suggest a margin for effective resolution. When I tested Photoshop’s Preserve Details resampling algorithm by downsizing USAF 1951 HiRes test target images it could retain the effective pixel resolution when the target was tilted slightly and exceed the effective pixel resolution when the target was tilted by 45°. Only when the target was oriented in parallel to the pixel grid a very small margin of 1-2% was required to avoid aliasing and resolve the target resolution.

The other crop of the Minolta 5400ppi scan contains finer details and some higher contrast lines that spread just over a single line of pixels indicating that the effective resolution and level of real detail in that area is very close to the sampling resolution.

You know, I'm not sure it is. Most of the efforts to determine scanner resolution are based around an analog test chart. Even some of the claims as the terrible MTF of the Epson appear to be based on a fairly tortured paper that attempted to use the USAF target in a totally different way than it was designed, to calculate the MTF.

I realize the fundamental purpose of a scanner is take an analog image and turn it into a digital file-- and anytime you convert analog to digital or vice versa, things can get a little squirrely-- but I have difficulty believing that Epson could build a scanner, claim it's capable of 6400 PPI, and not get hit by the mother of all class-action lawsuits because it can barely manage 2400 PPI. If the scanner had come out last year, fine-- but the v700/v750 came out in 2006, at a time when film -> digital was all the rage.

Something doesn't add up. Even the original discussion in this thread, produced a large debate as to what "125 lp/mm" actually meant. Silverfast, using their algorithm, suggests that that translates to 3175 PPI. I believe someone in this thread suggested that actually, 6350 PPI would be required, per Nyquist (while shaky on frequency sampling in general, I had the impression that the 125 lp/mm measurement was already taking Nyquist into account, and therefore 3175 would be sufficient).

Much of this feels like a great deal of hand-waving and armchair engineering-- the sort that suggested that you couldn't breathe in a car going 35 mph, or that bumblebees don't have enough lift to fly-- all great theories, but if empirical evidence contradicts the theory, a new theory is required. I'm not dismissing the information in this thread-- I think there's some really solid information, and I do appreciate all the contributions (and if I offended anyone by trying to steer away from hardware, that was not my intent. The thread on LF forums got pretty ugly at one point).

But the basic fact is that people are getting quality scans from the Epson scanners. They can get better ones from a drum scan (very expensive, very fiddly, and equipment is getting scarce), or from a DSLR (I suspect the quality of the lens is a factor here-- but can be *very* fiddly if you want better resolution than the Epson, and the price is either comparable at the low end, or in the "how high is up" category at the X1 end). In theory, dedicated film scanners such as the PlusTek should deliver better results, but no unit since the discontinued Coolscan has received consistently good reviews-- so it's expensive *and* obsolete.

The 125 lp/mm value is line pairs per mm. You need at least two pixel rows per line pair. Therefore, the 6350 ppi minimum requirement to resolve 125 lp/mm is correct.

Resolution test charts such as the USAF 1951 contain high-contrast line patterns which are useful to determine a scanner’s resolution limit as those high-contrast lines can still be resolved when the scanner's MTF is very low. When digitizing photographic film, the contrast of the highest frequency details is extremely low and thereby requires a much higher MTF from the scanner to transfer that level of detail into the scan.

My ICG 370HS drum scanner can resolve the 112 lp/mm line pattern parallel to the pixel grid of a high-contrast target (Danes Picta FSR1) and the 125 lp/mm line pattern at an angle of 45° to the pixel grid. As the drum scanner’s MTF at those frequencies is extremely low its performance for real world film digitization is only in the 80-100 lp/mm range, depending on the actual contrast of the detail on the film. Only with Adox CMS 20 my drum scanner’s performance is comparable to the high contrast test chart result. But that resolution is not even half of what Adox CMS 20 is capable of. Therefore, drum scanning is not adequate for digitizing Adox CMS 20.

So, for high-resolution film scanning the scanner must have sufficient MTF at the intended target resolution to be able to resolve the finest details of the film. The MTF of a scanner can be measured by scanning a slanted edge target and then analyzing that slanted edge by a software such as MTF Mapper. A target that has slanted edge features at the center and all four corners of the scanning area is useful for determining performance falloff towards the corners or misalignment issues. A high-quality USAF 1951 target can be utilized as a slanted edge target when scanned slightly tilted (about 5-6° is recommended).

Perceived scanner performance is often a matter of what references you have for comparison. For a long time, the drum scans I could get from my ICG had been the gold standard for me, though I was aware about its limitations. It was only after looking into the high-end of modern camera-based scanning that I learned to overcome those limitations and develop a system that is even adequate for lossless digitization of Adox CMS 20. On this super-high-resolution film the difference is obvious, but I was surprised that my new system could even extract slightly more details from an 8x10” transparency than I had previously drum scanned at 4200+ ppi.

Another aspect to consider is how well the scanner performance holds up across the density range. CCD and CMOS based scanners may suffer from sensor noise or banding in the higher density regions which also affects resolution. As my drum scanner, and I believe most drum scanners do, sends the PMT-signal through a log amp before A/D conversion the signal is boosted in the higher density tones for better shadow detail rendition but at the expense of tonal compression in the 3/4 tones where it has noticeably weaker MTF. It’s not really obvious unless pointed to and compared against a scan that doesn’t have this issue. And even then, when I made comparisons for clients or sent over demo files, some actually still prefer a drum scan for its own characteristic tonal rendition.

 

[Les Sarile]

First of the Epsons can provide a good scan. This one from a WWII 4X5 taken by a colleagues dad who was a combat photographer in Europe. Laid directly on the glass. No post work done other then cleaning up some dust and scratches.

Full res 2400dpi file -> http://www.fototime.com/5DCEE0B3D75F78C/orig.jpg
Couldn't have done it with my Coolscans and my colleague liked it well enough to get his own V700 to scan his dad's films.


However, don't you wonder why there is no improvement at higher dpi settings?
This one of Fuji RVP with the Epson V500, No improvement past 2400dpi.

Full res file -> http://www.fototime.com/33269E445D10043/orig.jpg

This one using the V700 of the same frame of Fuji RVP. No improvement past 4800dpi.

Full res file -> http://www.fototime.com/11F59FA46FF9497/orig.jpg

But Epson's 4800dpi - or 6400dpi, does not resolve more then Coolscan's 4000dpi. It doesn't even match it. Instead it is far lower. Here is the same frame of Fuji RVP using the Coolscan.

Full res file -> http://www.fototime.com/02BB797801DCA89/orig.jpg

It's unfortunate that there is no standard "real world test" frame of film we could all use. Individually we have to decide for ourselves what is good enough for our own real world results.

 

[138S]

This one using the V700 of the same frame of Fuji RVP. No improvement past 4800dpi.
...
But Epson's 4800dpi - or 6400dpi, does not resolve more then Coolscan's 4000dpi. It doesn't even match it.

I fully agree, It may make sense to scan 4800 or 6400 with the Epson a 35mm shot, to a allow a more refined sharpening, but of course resolving power is limited to 2900/2300 H/V, that's pretty clear. Not extrange that the Nikon can nail near 4000 effective at 4000 sampling because the nikon covers a single strip while the Epson lens is covering 4 strips, the nikon is a superior system for roll film, dot.

Anyway, from your V700 scan there are several questions. What holder ? What focus (holder height) optimization ?

If you used the original V700 holders focus was not specifically tunned to your particular unit. The new ANR holders ensure flatness and allow to adjust height in the four corners.

In that 2019 test, that OP used 3rd party adjustable holders and he nailed focus with a 1951 glass slide, a proficient job, or course if not having done that he would have not matched the result from the scanmate and from the creos.

The question is if a well made V700 Velvia scan would be much different from the Nikon one if the V700 scan was made proficiently with a good holder and a refined focus tunning. Velvia resolves this in practice:

Those numbers are debatable in (say) +/-10% , but the reality is not very far from these points. If this is true then a better scanner than Epson won't make much a difference even in a lab test, again if flatness/focus nailed.

Once you go to real photography, with things in the DOF and perhaps handheld then not many shots will surpase the Epson capability, may be 5% of TMX shots would deliver an slightly noticeable benefit from the Nikon superiority (over a V700 scan with focus/flatness nailed), and this would only be seen well beyond x12 where pictorial quality is usually at challenge. Yes... TMX on tripod and good glass/aperture may resolve 90 lp/mm, but at x8 you only see 48lp/mm of the negative in the print. Want quality ? Shot MF or LF !

The point is what X enlargement will show a benefit from an scanner that's better than the Epson.

My personal findings are next:

> x14 if flatness and focus are nailed in the Epson, proficient operator.

> x8 with the antique holders with a focus miss of 2.5mm, film was mostly flat

> x4 with curled film, plain focuss miss and an a clumsy operator.

 

[138S]

While the Shannon-Nyquist sampling theorem also applies to downsizing images, it does not necessarily suggest a margin for effective resolution.

This is not I was told when I was an student...

https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem

Of course reconstruction depends on the nature of the function, but an image is a "worst case" because the function is totally arbitrary, still we may consider that our function has a Fourier Transform, to apply that theory body. That "worst case" scenario implies that you need that safety margin in the samplling rate to not loss too much.

If you downsize to 2600dpi a larger image that is 2600 effective you will lose the 2600 performance. Try to sample a 20 line pairs (40lines) with 40 pixels, you'll get pure aliasing !!!

The other crop of the Minolta 5400ppi scan contains finer details and some higher contrast lines that spread just over a single line of pixels indicating that the effective resolution and level of real detail in that area is very close to the sampling resolution.

Again, from a super pixel-peeping, see how many pixels an edge takes for the transition, this not worth the half, clearly under 2600.

 

[Team ADOX]

There are several things that are highly problematic with those articles.
.............
Clarke has a few examples from a drum scanner. But we know nothing of it’s calibration or other circumstances concerning the scanning.
With that out of the way, the articles themselves are technically and semantically dubious in a number of ways.
There is much weasel wording and handwaving, where the authors think they can get away with it.
There are no concise parts part about method, or any rigorous references to other works or sources to speak of.

These are not scientific or academic peer reviewable articles (and they don't purport to be), though I guess for many people they give off an academic air, that impresses and appears to have special authority, due to the authors clearly being used to writing academic papers.

Tim Vitales article while interesting in a number of ways, appears to be mostly a scrape of other papers and articles, with little work done by the author himself.
Tim Vitale is a conservator who specialises in photos. That does not tell us much about his technical authority.

Clarkes article has a number of gross inaccuracies, omissions, lack of clarity, forces conclusions and frankly seems biased (perhaps for the reasons I mentioned to begin with).

That wouldn’t be such a problem if the two articles weren’t among the first that pops up with a google search, and if they didn’t still get cited time and time again.

We can confirm that from our own countless resolution tests over the years.
The articles of Clarke and Tim Vitale unfortunately have many errors and give overall a wrong assessment of the performance of film. We cannot recommend them.

There have been several examples or hints at just how much resolution is in a square inch/mm of film in the intervening fifteen years.
Here is a two examples;
Tim Parkins tests are pretty famous.

And they deserve it. The work of Tim Parkin, Henning Serger and Carl Zeiss belong by far to the best of what has been published so far concerning that topic.
We can confirm their results from our own work for years in that area.
ADOX is offering the highest resolving photographic medium for many years now (CMS 20 / CMS 20 II). And with HR-50 / Scala 50 we have another film in our product portfolio which offers very high resolution.
And of course a huge amount of tests is needed to evaluate such materials and their performance. Including all the imaging chains our (potential) customers would probably use.
For the best possible evaluation of film resolution we are using a microscope. With that you really see what incredible detail is recorded by excellent films.
Then further tests are made with classic optical enlargements: We've found that APO enlarging lenses have an outstanding performance, very close to microscope results. Users of negative film get the best performance using this workflow. It's surpassing even drumscanners by a very significant margin.
The same is true for slide projection: Performance with the best projection lenses is about on the same level as the APO enlarging lenses. As we are offering two BW reversal films we've done lots of tests in this field as well.
With these two workflows film users get the best resolution values for the final pictures. We highly recommend them. Also because you get the best performance by very low costs (often lower compared to scanners).

The second best performance concerning resolution are offering drumscanners. Really very good performance, sufficient for huge prints, exhibition and gallery work. Drumscanners are unmatched in the field of shadow and highlight detail capture and recovery (e.g. if you have done a mistake in exposure). That is due to their PMT technology.
We have some drumscan examples on our website, e.g some made by the leading German drumscan company www.high-end-scans.de. The expert behind this company, Mr. Ventzke, has an excellent reputation.

After that, on the next lower resolution level you'll find scanners like the last Nikon Coolscan series (5000; 9000), the Reflecta RPS 10M (and his Prime Scan sibling), and minilab scanners like the Noritsu HS-1800. There are some smaller resolution differences between them, but not huge ones.
We will probably also do some tests in the coming months with the new Plustek Opticfilm 120 Pro.

Next lower resolution level are scanners like the Reflecta MF 5000 (multi-format scanner) or the Plustek 8200 (35mm).

And at the bottom, with the by far worst detail resolution performance, there are the flatbed scanners.

We've done a huge amount of different tests with all these options in the imaging chains over the years (and continue to do so). And the ranking above is the result of these numerous tests.

ADOX - Innovation in Analog Photography.

 

[Team ADOX]

Velvia resolves this in practice:

View attachment 253324

That is absolutely wrong and has nothing to do with reality. The author of that
- has no knowledge at all of lens performance
- has never done resolution tests by himself.
Because the lens resolution is much much higher than that (and therefore also the resulting system resolution). For many decades now good prime lenses are either only diffraction limited, or are very close to that at their sweet spot.

We are seeing here again a typical internet problem: People, who have never done such tests by themselves, are quoting other people who have never done such tests by themselves.
And that results in wrong and misleading statements.

ADOX - Innovation in Analog Photography.

 

[Team ADOX]

The resolving power of most films is much higher than you think. Don’t trust the calculated resolution figures in Tim Vitale's document that you quoted. His calculations are fundamentally flawed. A much better source are the film resolution test results of Henning Serger. JP Buffington picked up some of Henning’s posts on photrio in his blog: http://jpbuffington.com/?p=167

Henning published additional results in other forums as well. Having personally checked a few of his test films under the microscope at 100x magnification I can confirm the credibility of his results. When I did my own tests in 2008 and 2009 the results were very similar. As these are real world tests of system resolution, Henning’s results can be taken as a reference on the potential resolving power of the emulsions he tested. While he used very good taking lenses (Zeiss Makro-Planar 2/50 and Nikkor 1.8/50 AI-S) for his tests these lenses are not the cutting edge by today’s standards. More recent introductions of new lens designs will probably yield even higher resolution from 35mm film.

Absolutely correct.
We can confirm your assessment both from our own numerous test results, but also from real life checks of the results of test expert Mr. Serger.
He has had countless meetings with several film and lens manufacturers and presented his test results in detail at all these meetings. He has also been at our company, presented his results to our CEO and our specialists, and we've had a very close and critical look at his results several times. We really appreciate his work. It is outstanding. He is a very reliable and trustworthy source. He is active in this area for about 30 years now. And he was able to give us even some suggestions and tips to further improve our own test methods.

ADOX - Innovation in Analog Photography.

 

[138S]

That is absolutely wrong and has nothing to do with reality. The author of that
- has no knowledge at all of lens performance
- has never done resolution tests by himself.
Because the lens resolution is much much higher than that (and therefore also the resulting system resolution). For many decades now good prime lenses are either only diffraction limited, or are very close to that at their sweet spot.

We are seeing here again a typical internet problem: People, who have never done such tests by themselves, are quoting other people who have never done such tests by themselves.
And that results in wrong and misleading statements.

ADOX - Innovation in Analog Photography.

Me, I test my lenses
___________________________________


Yes... lenses may approach to diffraction limit, but when they are near to that they are also at contrast extintion, and at contrast extintion film losses most of its resolving power.

TMX is rated 200lp/mm in the datasheet, but this is at TOC 1000:1 contrast, by no means you will find a pictorial situation in what you project 1000:1at 200cy/mm on film, at 200lp/mm you project zero contrast.

Also 1000:1 contrast is 10 stops, what you see in the test is the behaviour of the ultra low speed cubic layer that lays under the TMX main emulsion, but not the main emulsion that makes 99% or the pictorial work. In practice most of the image is at TOC 1.6:1 (65lp/mm before lens degradation), some edges may be 3 or 8:1, you may need an strong backlight silhouette in perfect focus and on tripod to get a ideal situation for high film performance, but this is a very particular situation. A contrasty very, very black line on a perfect white background is to deliver 8:1, and usually things are more or less gray than totally black or white.

Of course CMS 20 microfilm is another beast, monodisperse ultra low ISO boosted to 12 or 20. That fims records exactly what lens delivers... but it's a pictorial usage of microfilm rather than a pictorial film.


Then lenses don't yield it's peak performance by far in practical photography because real subjects are in a 3D world and things are in the DOF and scarcely in the perfect plane of focus, and even we may shot handheld, and ideal aperture may not be suitable or possible.

There is a way a subject is in perfect focus: a distant mountain, all at infinite focus, but even in that case atmospheric haze degradates the image to not have high contrast at high cycles.


______

I totally agree with the performance scale of scanners, just a discrepance...

The plustek 120 and the consumer 7000 / 8000 yield the same:

https://www.filmscanner.info/en/PlustekOpticFilm7500i.html
https://www.filmscanner.info/en/PlustekOpticFilm120.html

The 35mm units have an easier job, as only 1" is covered, while the 120 has to cover 6cm

 

[Lachlan Young]

There's no evidence that T-Max 100 has the same structural layout as T-Max 400. Indeed, given that T-Max 100 was available as glass plates until quite late on, all the evidence points to a blended set of emulsions coated in a single layer.

 

[Alan Edward Klein]

First of the Epsons can provide a good scan. This one from a WWII 4X5 taken by a colleagues dad who was a combat photographer in Europe. Laid directly on the glass. No post work done other then cleaning up some dust and scratches.

Full res 2400dpi file -> http://www.fototime.com/5DCEE0B3D75F78C/orig.jpg
Couldn't have done it with my Coolscans and my colleague liked it well enough to get his own V700 to scan his dad's films....
Full res file -> http://www.fototime.com/02BB797801DCA89/orig.jpg

It's unfortunate that there is no standard "real world test" frame of film we could all use. Individually we have to decide for ourselves what is good enough for our own real world results.

That's a terrific scan Les. How did you do it? What settings?

 

[138S]

There's no evidence that T-Max 100 has the same structural layout as T-Max 400. Indeed, given that T-Max 100 was available as glass plates until quite late on, all the evidence points to a blended set of emulsions coated in a single layer.

Shanebrook pages 18 and 20 shows that both TMY and TX are made coated in two sensitive layers, nothing is said about TMX.

What evidence is there about TMX having single layer ? Look at page 18 (figure 16, TMY), if cubic crystals were blended in the T layer then this would destroy the perfect horizontal arrangement of the flat crystals, consider this, it would be painful destroying that.

Anyway this is not relevant for that reasoning, at TOC 1000:1 there is a 10 stops range, we see resolution from ultra slow speed crystals overexposed +7 stops at least, as we have some 3 stops from meter point to toe. Those crystals would be under "equivalent" ISO 1.

 

[Lachlan Young]

Shanebrook pages 18 and 20 shows that both TMY and TX are made coated in two sensitive layers, nothing is said about TMX.

What evidence is there about TMX having single layer ? Look at page 18 (figure 16), cubic crystals blended in the T layer would destroy the perfect horizontal arrangement of the flat crystals, consider this, it would be painful destroying that.

Anyway this is not relevant for that reasoning, at TOC 1000:1 there is a 10 stops range, we see resolution from ultra slow speed crystals overexposed +7 stops at least, as we have some 3 stops from meter point to toe. Those crystals would be under "equivalent" ISO 1.

If you'd read the segment on plate coating you'd know that you would not be able to coat a supercoating layer on the machine if T-Max 100 was multilayer. The gap between single and multilayer coated BW films often falls somewhere above 100 and below 200-250 as the problems of single layer blended coatings increase to where they are more problematic than the disadvantages of multilayer. As far as is known, T-Max 100 consists almost entirely of flat/ tabular crystal structures (of such great available surface area that the supercoat has to be unusually thick to moderate development times) - the two different layers seems to have been an innovation borrowed from colour neg technology for T-Max 400. Pretty much no 100 speed BW film on the market is multilayer as far as has been disclosed, though they almost all use some degree or other of emulsion blending. If you tried to use the structure shown in T-Max 400's cross section as a single layer, you would likely run into equilibration issues. There is often a significant coated thickness difference between 100 and 400 speed films & this is at least in part due to the multiple layers in the faster film. But this not surprising given the way that you start from the perspective that you believe that you are more knowledgeable than the entirety of Ilford/ Kodak/ Agfa/ Fuji/ Adox's R&D...

 

[Team ADOX]

TMX is rated 200lp/mm in the datasheet, but this is at TOC 1000:1 contrast, by no means you will find a pictorial situation in what you project 1000:1at 200cy/mm on film, at 200lp/mm you project zero contrast.

Please, this permanent concentration on and "fetish" of these data sheet numbers really lead you in the wrong direction:
The behaviour of resolution in its dependency on object contrast is not completely linear over the whole range from 1.6:1 to 1000:1. That is very important to know!!
To put it in other words and make it more clear: To come close to the resolution value of the 10 stops (1000:1) of 200 lp/mm you don't need 8 or 9 stops difference, but only 4-5 stops.
For example Carl Zeiss has achieved 180 lp/mm resolution with TMX with that 4-5 stop contrast. And such a contrast is not seldom in normal shooting conditions.
We have seen TMX examples with resolution of about 135 lp/mm at an object contrast of only 4:1 = two stops (that were results by Mr. Serger).
And both the Zeiss and Serger results were achieved with normal equipment which is available for every photographer.
So we have the "law of diminishing returns" here: From 1.6:1 to 16:1 or 32:1 you have a quite linear response, and with even higher object contrasts the curve is significantly flattening.
The resolution is quite high already in the low(er) to low-medium contrast range.

In practice most of the image is at TOC 1.6:1 (65lp/mm before lens degradation), some edges may be 3 or 8:1, you may need an strong backlight silhouette in perfect focus and on tripod to get a ideal situation for high film performance, but this is a very particular situation. A contrasty very, very black line on a perfect white background is to deliver 8:1, and usually things are more or less gray than totally black or white.

No, you don't need such situations. Please just make the following: Go out with a camera with built-in spot meter. And then meter the contrast difference of details. For example the name plate of a brick-and-mortar shop, name in contrast to the underground. Or traffic signs: The picture or text to the underground. Or street signs: Name text in relation to its underground. Make the same with licence plates / number plates of cars. Window frames of houses in contrast to the house walls, name plates on historic buildings......etc (that are details you want to be resolved in pictures in most of the times).
You will often meter 2-3 stops contrast difference in such cases. In even, dull (overcast) light. No backlight contrast at all.
And at such low to moderate contrasts films like for example HR-50, TMX, Delta 100, Provia 100F, Velvia 50 / 100 etc. achieve resolution values in the 120 to 150 lp/mm range (depending on the used lens).

ADOX - Innovation in Analog Photography.

 

[Alan Edward Klein]

If you'd read the segment on plate coating you'd know that you would not be able to coat a supercoating layer on the machine if T-Max 100 was multilayer. The gap between single and multilayer coated BW films often falls somewhere above 100 and below 200-250 as the problems of single layer blended coatings increase to where they are more problematic than the disadvantages of multilayer. As far as is known, T-Max 100 consists almost entirely of flat/ tabular crystal structures (of such great available surface area that the supercoat has to be unusually thick to moderate development times) - the two different layers seems to have been an innovation borrowed from colour neg technology for T-Max 400. Pretty much no 100 speed BW film on the market is multilayer as far as has been disclosed, though they almost all use some degree or other of emulsion blending. If you tried to use the structure shown in T-Max 400's cross section as a single layer, you would likely run into equilibration issues. There is often a significant coated thickness difference between 100 and 400 speed films & this is at least in part due to the multiple layers in the faster film. But this not surprising given the way that you start from the perspective that you believe that you are more knowledgeable than the entirety of Ilford/ Kodak/ Agfa/ Fuji/ Adox's R&D...

I use Tmax 100 in MF and 4x5. What advantage do the extra tabular layers and other criteria of this film that make this film different? What practical effect would I see in my photography?

 

[Lachlan Young]

I use Tmax 100 in MF and 4x5. What advantage do the extra tabular layers and other criteria of this film that make this film different? What practical effect would I see in my photography?

The thin single-layer, multiple emulsion type of construction allows for inherently very high sharpness (less internal reflection between layers) with a useful range of characteristic curve behaviour and the flat T-grains allow for a potentially much better speed / grain relationship with grains more tightly packed for better coverage - at the cost of a possible granularity character that some seem to spend a lot of time obsessing over. Not really an issue I've found. Faster films have to use more complex techniques to improve speed/ grain & combat internal reflections from consequent multi-layer approaches. The highly efficient dye sensitising techniques used by Kodak add a further chunk of speed & are almost as important as the grain structures.

 

[Alan Edward Klein]

The thin single-layer, multiple emulsion type of construction allows for very high sharpness (less internal reflection between layers) with good latitude and the flat T-grains allow for a potentially much better speed / grain relationship with grains more tightly packed for better coverage - at the cost of a possible granularity character that some seem to spend a lot of time obsessing over. Not really an issue I've found. Faster films have to use more complex techniques to improve speed/ grain & combat internal reflections from consequent multi-layer approaches. The highly efficient dye sensitising techniques used by Kodak add a further chunk of speed & are almost as important as the grain structures.

How would these features show up in my photography? What would it look different to a person who didn't know what film I used?