Film vs. Scanning resolution

[Team ADOX]

Also think that even a superb lens kills most of the contrast at high cycles.

Wrong.
If that would be the case generally, all the new, modern and improved lenses we have (fortunately) seen in the last 30 years would not make much sense.
Look at the classic car examples on our homepage (the MG and Porsche). You see even the tiniest little dirt spots, really very low contrast details. Because lens and film can record them.
Current very good prime lenses transfer 60-80% of contrast at 40 cycl./mm. Excellent results.
There has been lots of progress in lens design in the last decades.

ADOX - Innovation in Analog Photography.

 

[Lachlan Young]

For film, MTF graphs are for an specific contrast ratio, this is not the same case than with lenses.

With film, depending on the contrast ratio grains of different sizes take action, so depending on contrast (and base exposure) the MTF graph is totally different.

In the 1000:1 contrast there is a 10 stops difference between lines, so in practice this can be made with a contact copy or perhaps with something like a semiconductor integrated circuit manufacturing lens (projection with reduction). By controlled fogging the film in advance (or after) you adjust contrast of the test and the base exposure of the (projected) black lines.

In that 1000:1 test, if black (projected) lines are in the toe (or not exposed) then white lines will be overexposed around +7, so you are to detect the effect of the ultra small grains having high resolving power over a "fog" from more sensitive crystals... 1000:1 does not describe the film behaviour in pictorial situations as you test crystals that are sensitive at +7 overexposure.

Instead a 1.61:1 tells the truth for common textures, 8:1 for contrasty edges, and 30:1 is good to guess performance for backlighted silhouettes.

If you see the TMX datasheet it says Modulation Transfer extintion at 200 lp/mm for TOC 1000:1 and 65lp/mm for TOC 1.61:1

The posted Provia graph is the one for the unobtanium 1000:1, clearly it has a tendence to extintion at 160lp/mm, well... commercially this MTF graph is nicer than the 1.61:1 would be.

I llustrated in green what it could be the 1.6:1 MTF graph, with extintion at 60lp/mm in that case.

This is total and utter nonsense that displays a complete lack of even the most basic comprehension of MTF and massive confusion with resolution testing. Film MTF explains how much of the contrast of the target exposed to it will be replicated: ie at 50 cyc/mm T-Max 100 should reproduce 100% of the contrast in the target, at about 130cyc/mm it'll reproduce 50% of the contrast of the target, and at 10cyc/mm it'll reproduce about 120% of the target's contrast - which will make it look sharper than it is in reality. The actual contrast of the target is relatively immaterial, it's about how much of that target's contrast is able to be captured by the film at a given resolution. The number of data points gathered & methodology used to arrive at those results published by Kodak, Fuji etc are vastly larger and more complex than you want to deceive everyone into believing. MTF tests are quite specifically (and rigorously) carried out to ensure that they deliver a fair representation of how the material will behave in everyday usage.

It's the same with scanners - if a system can only deliver 50% of the contrast of any given target or piece of film at 13 cyc/mm, it is going to simply fail to record a huge amount of critical object sharpness information that a scanner that can deliver 50% target contrast transfer at 25 cyc/mm or even 50 cyc/mm will readily pick up.

Lens MTF charts show contrast performance at lower cyc/mm for various reasons, but the point of their MTF charts are to show the relative performance across the field. For obvious reasons, if film MTF behaviour varies in a statistically significant way across the width of a coating batch, you have a serious QC problem on your hands. It is now known that very high MTF at low cyc/mm is essential to a lens appearing 'sharp' rather than outright 'resolution' - otherwise a disposable camera lens would be outperforming the various expensive LF lenses at f32 or deeper stops.

It is very obvious that your 'experience' is so limited in qualitative scope by low-grade techniques/ equipment/ usage of equipment that you are desperately trying to find excuses/ clickbait that exculpates those self-same pieces of equipment/ techniques etc.

 

[Helge]

8000ppi with a 24 MP FF DSLR as a scanner? No, definitely not. We've never seen such results. It is physically impossible, as the Nyquist frequency of these sensors is too low.

ADOX - Innovation in Analog Photography.

With extended bellows and an extension ring or two (and stitching of course) I’d say I can hit 8000 dpi with my 24mp DSLR..
I have never tested it rigorously though.
But I do get a clear view of the grain.
You do need a strong backlight to focus though because of the immense light loss.

 

[138S]

Current very good prime lenses transfer 60-80% of contrast at 40 cycl./mm. Excellent results.

Of course... yes at 40cy/mm...

But when when going to high cy/mm the thing is different. That $2800 Zeiss, by 120 lp/mm, has a 50% contrast loss in the center and a 100% in the mid-corner.

 

[faberryman]

Anybody changed their mind in the last eleven pages of discussion because it sure looks like a merry-go-round to me. I shot film and scanned it for a little while and wasn't satisfied with the results. I now shoot digital and it is mostly black and white. You might want to give it a try.

 

[138S]

This is total and utter nonsense that displays a complete lack of even the most basic comprehension of MTF and massive confusion with resolution testing. .

Lachlan, your personal attacks are boring

Read ISO 6328:2000, try to understand it, and you won't say silly/ridiculous things.

For each contrast ratio and base exposure you have a different MTF graph for the film, MTF response changes with base exposure and with contrast.

I shot film and scanned it for a little while and wasn't satisfied with the results. I now shoot digital and it is mostly black and white. You might want to give it a try.

I shot digital and processed it in Photoshop for a little while and wasn't satisfied with the results. I now shoot film, it is mostly black and white and color, I print BW in the darkroom from 35mm to 5x7", and contact print 8x10".

In the darkroom tonality is quite harder to control than with Ps, but a big print from 5x7" optically crafted on silver paper is another league, still what counts is what the image says, of course.

You might want to give it a try.

YMMV.

 

[warden]

It's a shame that Zeiss hasn't continued to support Camera Lens News archives. There was so much good content there that is still relevant.

 

[Lachlan Young]

Read ISO 6328:2000, try to understand it, and you won't say silly/ridiculous things.

Resolving power ≠ MTF. Do you understand that? MTF is the percentage of target contrast reproduction at a given resolution. MTF test target contrasts are (and should be) controlled for, so that they are irrelevant to the final results. Do you understand this? MTF and RP have a relationship, but it is not the linearity you are erroneously assuming it is.

 

[138S]

Resolving power ≠ MTF. Do you understand that? MTF is the percentage of target contrast reproduction at a given resolution. The target contrasts are (and should be) controlled for. Do you understand this? MTF and RP have a relationship, but it is not the linearity you are erroneously assuming it is.

Finally, I guess I found the way to make you understand it:

For a lens you have a MTF curve for each aperture, sure you understand this...

Well, for film it happens something similar, you have a MTF curve for each contrast.

For a lens what changes the curve is the aperture, for film what changes the curve is contrast. Do you understand it right now?

Question:
Why, with film, the modulation transfer varies with contrast ?

Answer:
Because the kinds of crystals that are exposed are different, like if was another film type !!!!!

Did you catch it ? Don't tell me you were not aware...

 

[Lachlan Young]

Well, for film it happens something similar, you have a MTF curve for each contrast.

No. This is completely and utterly wrong. You have a single MTF curve for a BW film and 3 (or an average, though usually the green is chosen) for each of RGB sensitive layer groups in a colour film. The absolute target contrast has nothing to do with MTF - MTF is about how well the contrast of any target is reproduced at a given resolution. Testing is carried out using X-rays and gratings of specific micron widths, both positive and negative, and read on a microdensitometer, then plotted. Target contrast matters to Resolving Power and not to MTF.

 

[SCHWARZZEIT]

For clarification:
In their Professional Film Guide Fujifilm gives a pretty detailed description on how both MTF and the two resolving power values are measured: https://asset.fujifilm.com/www/us/f...4c91bcedb7de050/ProfessionalFilmDataGuide.pdf
See pages 123-126 and pages 128-129.

For MTF measurement they use sinusoidal frequency patterns of varying density and contrast. The document doesn’t state which contrast ratios they test. But the MTF represented in the data sheet is normalized for the contrast range that their test procedure covers.

To measure resolution values, they take photographic images of a high-contrast test chart and low-contrast test chart with the film in camera utilizing a high-resolution lens. There is no information on which lens or aperture is used, but apparently, at least the diffraction-based reduction of contrast is already accounted for in Fuji’s film resolution specifications. These number are very close to what other independent testers have found which is good evidence that they reflect the real-world resolution potential of these films with high-resolution lenses.

 

[Lachlan Young]

For MTF measurement they use sinusoidal frequency patterns of varying density and contrast. The document doesn’t state which contrast ratios they test. But the MTF represented in the data sheet is normalized for the contrast range that their test procedure covers.

That's exactly what I've read from other sources too - and have seen the mathematical steps used to arrive at that point from the raw datasets.

 

[Helge]

Anybody changed their mind in the last eleven pages of discussion because it sure looks like a merry-go-round to me. I shot film and scanned it for a little while and wasn't satisfied with the results. I now shoot digital and it is mostly black and white. You might want to give it a try.

A true scholar has spoken.
I’m sure you can print some nice postcards.

BTW what happens to tonality and resolution when you need to extract pure R, G or B for filter simulation?

 

[Alan Edward Klein]

Anybody changed their mind in the last eleven pages of discussion because it sure looks like a merry-go-round to me. I shot film and scanned it for a little while and wasn't satisfied with the results. I now shoot digital and it is mostly black and white. You might want to give it a try.

Frank You have nice work on your site. Can you identify which were done in digital vs. film?

 

[138S]

The document doesn’t state which contrast ratios they test. But the MTF represented in the data sheet is normalized for the contrast range that their test procedure covers.

I guess you will agree that depending on contrast the curve will be one or another...

The MTF represented in the datasheets is always the 1000:1 one, it always hits the rated TOC 1000:1 resolution at extintion.

See also where it has to hit the graph at 1.6:1 contrast: 50cy/mm in 0% (at extintion).

Also realize that the vertical scale is Logarithmic, so it results more nice, but see the 10% transfer horizontal how it cuts the curves:

This was TMY.

 

[Lachlan Young]

I guess you will agree that depending on contrast the curve will be one or another...

The MTF represented in the datasheets is always the 1000:1 one, it always hits the rated TOC 1000:1 resolution at extintion.

See also where it has to hit the graph at 1.6:1 contrast: 50cy/mm in 0% (at extintion).

Also realize that the vertical scale is Logarithmic, so it results more nice, but see the 10% transfer horizontal how it cuts the curves:

View attachment 253798

This was TMY.

You have not read a word of that Fuji manual that SCHWARZZEIT linked you to. It explains in excellent and clear detail what the differences between MTF and Resolving Power are, and has pictures of the different types of charts used.

 

[grat]

8000ppi with a 24 MP FF DSLR as a scanner? No, definitely not. We've never seen such results. It is physically impossible, as the Nyquist frequency of these sensors is too low.

At the risk of interrupting Lachlan and 138S (again), let's see if my crash course in such things has been a waste of time... As I understand it, the Nyquist limit of a CMOS/CCD sensor is roughly equal to pixels/mm/2-- But then the bayer cfa (or equivalent) intervenes, and my brain explodes.

Regardless, it suggests that rather than a FF DSLR, a micro 4/3's would be better suited (I will not, nay, cannot, suggest a smartphone camera with a straight face. Even if it works, I will stick my fingers in my ear, and hum really loud)-- a FF DSLR with 24MP would have a limit of ~80 lp/mm, my 90D should be pushing 150 lp/mm, and an MFT (not MTF), with pixel shift, is just silly.

 

[SCHWARZZEIT]

I guess you will agree that depending on contrast the curve will be one or another...

The MTF represented in the datasheets is always the 1000:1 one, it always hits the rated TOC 1000:1 resolution at extintion.

The extinction resolution depends on the level of input contrast. The film MTF is only a description on how much of the input contrast is film recorded by the film. If we pick up your TMY MTF for a theoretical exercise and you start with 1000:1 contrast and the film has an MTF of 0.11% (extrapolated guess as Kodak does not provide data for this spatial frequency in the graph) at 200 lp/mm then the recorded contrast will be 1.1:1 and thus barely resolved detail. If your input contrast ratio is only 1.6:1 and the film MTF is 70% at 50 lp/mm then the recorded contrast on film will be also just about 1.1:1.

In practical photography it does not always work exactly like this as there are other factors to consider. The variably sized crystals reacting differently to light could be a factor. The MTF is only an approximation. Near a film’s resolution limit fine detail is rendered out of randomly distributed variable sized grain clumps and way too fuzzy to be accurately described by an MTF.

 

[138S]

At the risk of interrupting Lachlan and 138S (again),

OK, I won't answer Lachlan's anymore here.

 

[138S]

The extinction resolution depends on the level of input contrast. The film MTF is only a description on how much of the input contrast is film recorded by the film. If we pick up your TMY MTF for a theoretical exercise and you start with 1000:1 contrast and the film has an MTF of 0.11% (extrapolated guess as Kodak does not provide data for this spatial frequency in the graph) at 200 lp/mm then the recorded contrast will be 1.1:1 and thus barely resolved detail. If your input contrast ratio is only 1.6:1 and the film MTF is 70% at 50 lp/mm then the recorded contrast on film will be also just about 1.1:1.

In practical photography it does not always work exactly like this as there are other factors to consider. The variably sized crystals reacting differently to light could be a factor. The MTF is only an approximation. Near a film’s resolution limit fine detail is rendered out of randomly distributed variable sized grain clumps and way too fuzzy to be accurately described by an MTF.

Exactly.

But also consider that this 1000:1 or 1.6:1 is TOC, so it's the actual projected (or contacted) contrast on film, not the contrast of the physical target before projecting it. A real shot has the lens contrast transfer (at high lp/mm) limiting edge contrast and also flare decreases comtrast.

Also note that TOC 1.6:1 50lp/mm and TOC 1000:200lp/mm are at extintion, so conditions near 0% MTF response are reached.

this is totally well described in the ISO 6328 norm, that fuji document tells quite incomplete information in the critical points, but ISO 6328 is clear. I saw ISO 6328 doc several years ago, IIRC a limiting target element (extintion) was used to issue a rating.

For this reason the (TMY) MTF graph made at TOC 1000:0 has to hit coordinates [0%MTF , 200lp/mm] and the MTF graph made at TOC 1.6:1 has to hit coordinates [0%MTF , 50lp/mm] like the guessed lines I painted hit. Of course those lines are a theoric exercise, but they show perfectly the real nature of the situation.

Both kodak and fuji don't say the contrast in what their datasheet MTF is made, this is not nice. Actually there is no doubt because a curve tending to hit 200lp/mm at 0% MTF can only be the TOC 1000:1 one, but they should state it.

The MTF is only an approximation.

I agree... while the MTF chart itself is not an aproximation (it has to be the accurate measurement of the modulation transfer in the tested conditions) of course it does not say how much pictorially degradated the image is at high cycles for the contrast, as beyond the contrast loss we have to add grain noise to the degradation. Well, another way to say the same.

--------

Anyway we should be aware that the MTF charts in Fuji/Kodak datasheets are missleading, because they report a situation (on film TOC 1000:1) that is exeptional in real scenes, as a black line on a perfect white delivers 8:1 and contast fall in the projected edge is high cycles are considered.

--------

IMO, all that has to be well understood if wanting to undesrtand well what is required from scanners and in what situations. For example the Epson V700 (if well focused !) resolves 55lp/mm (48 in the Vert axis) at extintion, matching the extintion on film in many of the pictorial situations, for this reason the practical tests made by Alan and Pali show those matching results compared to high end machines, of course there are situations like the Serger's Porsche that are quite more challenging.

 

[Team ADOX]

With extended bellows and an extension ring or two (and stitching of course) I’d say I can hit 8000 dpi with my 24mp DSLR..
I have never tested it rigorously though.
But I do get a clear view of the grain.

Try it. But you certainly will not get your "8000 goal".

"But I do get a clear view of the grain".
But that does not mean at all that you get the full resolution. There is this myth on the internet, that when you can see the grain, you also have exploited full resolution.
But that is totally wrong.
Besides the fact that it is often more scanner noise or enhanced grain what you are seeing in scans, the max. resolution is a completely different animal:
We see that permanently in our test work: We can see clearly the film grain with a 20x grain focussing aid under the enlarger. But for evalution of the real, full resolution we have to use much higher enlargement factors: 100x under a microscope and slide projection.

ADOX - Innovation in Analog Photography.

 

[Lachlan Young]

Mainly because there seems to be a distinct effort afoot on the part of a particular individual to create a whole lot of misleading claims around sharpness/ MTF and Resolving Power, I'm posting screengrabs of the information pages from the Fuji document linked upthread. They make it clear what the differences are - and show the test charts used - though there is a slight typo in the Resolving Power segment.

 

[Helge]

OK, I won't answer Lachlan's anymore here.

I can see why.

Lachlan, this is what it looks like when you’ve nailed it.
You’ll never get any kind of admission out of philistines, only grumbling retreat, while they are shouting names at you.
A month later he will act like it was all obvious and “that’s what he thought all the time”.
In a year he will claim he discovered the idea.

 

[138S]

Try it. But you certainly will not get your "8000 goal".

"But I do get a clear view of the grain".
But that does not mean at all that you get the full resolution. There is this myth on the internet, that when you can see the grain, you also have exploited full resolution.
But that is totally wrong.
Besides the fact that it is often more scanner noise or enhanced grain what you are seeing in scans, the max. resolution is a completely different animal:
We see that permanently in our test work: We can see clearly the film grain with a 20x grain focussing aid under the enlarger. But for evalution of the real, full resolution we have to use much higher enlargement factors: 100x under a microscope and slide projection.

ADOX - Innovation in Analog Photography.

DSLR scanning can reach easily around 10000 dpi effective, by stitching a 3x3 or 4x4 mosaic from a 35mm frame. Element 7.5 of the USAF 1951 glass was resolved.

The way I found to do it was reversing an enlarger lens (Nikon EL 50mm), in that way focus in the film plane way is less critical, then the image is magnificated the image around x2.5 on the DSLR sensor.

Still a convenient setup should be made, but no doubt that by stitching crops enough we don't have practical limitations, but to overcome several problems a regular the lens may have to be reversed.

I realized that way while rating the performance of an Rodagon 210, I had the glass slide in the negative carrier and I was projecting it on the sensor of a DSLR facing up without the DSLR lens mounted, the slide was enlarged around x4 on the SLR sensor... the Rody 210 peaked 145lp/mm taken in the negative, real 35 lp/mm "on paper" at x4, I was just investigating what safety margin the Rodagon provided to not be a limiting factor in the print IQ...

 

[Helge]

Try it. But you certainly will not get your "8000 goal".

"But I do get a clear view of the grain".
But that does not mean at all that you get the full resolution. There is this myth on the internet, that when you can see the grain, you also have exploited full resolution.
But that is totally wrong.
Besides the fact that it is often more scanner noise or enhanced grain what you are seeing in scans, the max. resolution is a completely different animal:
We see that permanently in our test work: We can see clearly the film grain with a 20x grain focussing aid under the enlarger. But for evalution of the real, full resolution we have to use much higher enlargement factors: 100x under a microscope and slide projection.

ADOX - Innovation in Analog Photography.

I know, I know. I’ve been saying the same thing for years.
I had a feeling I should have corrected it, or worded it differently after I clicked post.
I mentioned grain to eschew the thought that we where talking diffraction limitation or something like that.
Seeing grain is an indication that you are at least focused at the surface of the film (which is not always what you want).

But grain is not in any way or shape resolution. In the same way tape hiss from magnetic grain in audiotape does indicate the upper frequency limit of the tape.

I can’t see why 8000 dpi wouldn’t be possible?
It’s certainly possible by putting the same camera on a microscope, which is in essence the same thing.
Or in extreme macro you can resolve the fine details of an insects eye, moss or stamen in the center of a flower.