Film vs. Scanning resolution

[Les Sarile]

I’ve read through that several times over the years. Quite a lot of interesting stuff.

Then you might also be interested in articles posted at https://clarkvision.com/articles/index.html
A good reference between film and digital with a lot of studies related to resolution.

 

[Adrian Bacon]

Then you might also be interested in articles posted at https://clarkvision.com/articles/index.html
A good reference between film and digital with a lot of studies related to resolution.

yep. Read a pretty good chunk of that too. Good stuff.

 

[138S]

Seemingly you post some nice references so how about the some of your own test results to show your point indisputably? Make it available for peer review.

My personal tests with the Epson consist in inspecting x60 (magnifier) the negative and comparing with the crop in the monitor with the negative original image to feel better any difference. I'd recommend this to anyone wanting to know if his scanner is doing a good job.

You can review the comparison at https://www.filmscanner.info/en/EpsonPerfectionV800Photo.html
Effectively, the V800 is equal to the V850 and achieves slightly higher res then the V700. The V800 has better holders and faster scan times too.

Let me add that the slightly better performance of the V800 over the V700 may be related to the ANR holders flattening better the film. The (cheap) V800 holders can be purchased to be used in the V700.

The V850 has an improvement over the V800 which is better coatings in the lens and in the mirrors, which would be useful to avoid flare in certain situations. Also the V850 bundled software features Multi-Exposure, with the V800 a SF software upgrade has to be purchased to have ME.

Anyway, to overcome flare and to get optimal performance with very dense slides (beyond using multiexposure), a nice trick is placing the film in the central receptacle (of the 35mm holder) and covering the other two holes with something opaque.

of course, if we place an ultra dense velvia slide in one receptacle and the other two are empty or holding very low density film then we are to add a lot of stray light to the system. Many times this is way more important than the better coatings.

Some reviews stating low DMax capability of the Epson were placing a dense media in a receptacle and leaving the others fully open, which is unfair, when the same people were recommending masking the film boundaries in the expensive high end and pre-press flatbeds.

 

[Lachlan Young]

If a scanner has a tested MTF that drops below 50% at under 14 cyc/mm, any contrast information on the film above 50% at 14 cyc/mm or more will be steadily more lossy in its transmission to the scanner. That certain films have 100+% MTF out to quite high cyc/mm helps compensate to some small extent, but the loss of contrast is going to be much greater than on a scanner/ camera sensor with much higher MTF. The rest is just somebody making excuses for the performance of a cheaply made flatbed scanner that they are obsessed with.

All the stuff about bigger dye clouds for scanning is completely wrong. Instead the grain has shrunk significantly, been more tightly packed and better sensitised enabling a better speed/ grain relationship & thus better detail rendering. If dye clouds were allowed to grow to the extent claimed they would reduce large object sharpness considerably. Current Portra films print very very well in traditional optical printing & the newer films are clearly better than the 1990's ones. The problem with older colour neg films was that it was possible to produce very strong edge effects/ microcontrast (helping with cheap amateur cameras with low contrast lenses) but at a significant noise cost. VR 100/ Gold 100 is as noisy as Portra 800 and this has consequences for the resolution of low contrast small objects, which get lost in the noise/ grain of the dye clouds. Adjusting the grain & sharpness slightly to reduce noise (and professional films are generally exposed with much higher MTF lenses) produces an overall better result. If you improve a film's darkroom performance, you will improve its scanning performance. A simple microscope will tell you very little about the actual structure of the material whereas optically printing it or scanning it on a high end scanner/ good camera scanning rig will tell you far more, much more accurately about the real-world performance of the material.

 

[138S]

If a scanner has a tested MTF that drops below 50% at under 14 cyc/mm, any contrast information on the film above 50% at 14 cyc/mm or more will be steadily more lossy in its transmission to the scanner.

You are a bit wrong, your Hasselblad X1 (when scanning 4x5") has not 50% MTF at 14lp/mm, it has 50% MTF at 15.75lp/mm. It is an small error but it is 15.57 and not 14

The X1 for 4x5" resolves 1575 dpi effective, in fact it matches its yield of 6300 for 35mm, 6300/4" is just 1575. This is total contrast extintion by 31.5 cy/mm and 50% MTF at 15.75cy/mm.



The Epson V700 resolves 55lp/mm (48 in the hor axis) at extintion from 35mm to 5.9" scan witdh (with ANR or equivalent holders), with a constant yield from the fixed lens that is always covering 5.9".

Instead the X1 varies its yield from the lens distance change (zoom in/out effect), delivering 126cy/mm at extintion for 35mm, 31.5cy/mm for 4x5" and 25cy/mm at extintion for 5x7". For 5x7" the X1 has 50% MTF at 12cy/mm.

The X1 is better for 35mm, the V700 is better for LF. For MF the X1 is slightly better but this is not much noticeable, if not stitching strips.

 

[Lachlan Young]

You are a bit wrong, your Hasselblad X1 (when scanning 4x5") has not 50% MTF at 14lp/mm, it has 50% MTF at 15.75lp/mm. It is an small error but it is 15.57 and not 14

The X1 for 4x5" resolves 1575 dpi effective, in fact it matches its yield of 6300 for 35mm, 6300/4" is just 1575. This is total contrast extintion by 31.5 cy/mm and 50% MTF at 15.75cy/mm.



The Epson V700 resolves 55lp/mm (48 in the hor axis) at extintion from 35mm to 5.9" scan witdh (with ANR or equivalent holders), with a constant yield from the fixed lens that is always covering 5.9".

Instead the X1 varies its yield from the lens distance change (zoom in/out effect), delivering 126cy/mm at extintion for 35mm, 31.5cy/mm for 4x5" and 25cy/mm at extintion for 5x7". For 5x7" the X1 has 50% MTF at 12cy/mm.

The X1 is better for 35mm, the V700 is better for LF. For MF the X1 is slightly better but this is not much noticeable, if not stitching strips.

And you ran an MTF test on an X1 when?

If you did, the onus is very firmly on you to show the raw results given your known reputation.

Or are you guessing wildly on the basis of no actual experience?

 

[warden]

So last night, I got to thinking about how much resolution is actually needed to scan film.

I haven't read through all the responses here so sorry if I'm repeating. If your question is theoretical rather than practical I have no answer for you and others here can dive into the math and lines per mm etc which is not my expertise.

But if you're thinking of practicalities I've found consumer scanners are fine for small prints (say 8x10-ish size) and can record the feeling of film in the image, but don't ask me to define that other than to say the prints look really good to me when I compare them to wet prints without using a loupe or a microscope or whatever. If I'm printing larger though I have to pay someone else to do the scanning because my consumer scanners (A Nikon Coolscan for 35mm and a V700 for MF) won't give enough information for a closer inspection of the paper, which people tend to do.

 

[138S]

And you ran an MTF test on an X1 when?

Yes, years ago I helped a friend to rate his X1 and his X5, by making high contrast contact copies of the 1951 slide on CMS 20, using microfiche developer.

Anyway those results are pretty consistent with all tests made around, https://www.filmscanner.info/es/HasselbladFlextightX1.html

https://www.filmscanner.info/en/HasselbladFlextightX1.html

Total contrast extintion at 6300 dpi effective at 1", which is contrast extintion at 1575 when zooming out x4 to catch 4" instead 1" with the same lens and sensor. Our tests delivered an exact inverse proportional yield to the scan width, which at all is not a surprise. Is it a surprise for you?

in fact, IIRC our test delivered an slightly lower average than the filmscanner.info result, anyway I used the 6300 value from filmscanner instead 6100 (average, IIRC) we got with the X1.

 

[grat]

no real need. This is already done in many image formats. The luminance information is stored in full resolution and the chroma information is stored in less resolution.

I believe you have shot right past my point.

I'm proposing a different methodology for scanning a color negative to produce an image format. What I got from the article was that the grain resolution is far higher than the color resolution, suggesting a two-pass approach to the scanning process.

 

[138S]

I believe you have shot right past my point.

I'm proposing a different methodology for scanning a color negative to produce an image format. What I got from the article was that the grain resolution is far higher than the color resolution, suggesting a two-pass approach to the scanning process.

Many scanners can do a multi-pass, it is a feature suported by SilverFast...

But in practice this won't be much worth.

 

[NortheastPhotographic]

As someone who scans 7 days a week for clients I can tell you I spend NO time thinking about MTF charts.

So many people get lost in techno jargon. Just set a target scan size that you need to achieve, and base that probably on the largest print size you're likely to make. When you make a scan, it should be roughly as detailed as you you expect it to be, with however much grain you also expect to see depending on the film. You can tell when your scanner is not up to the task (Epson) and when you may want to even doing a slight softening (Coolscan).

Lots of people will tell you that you need to have a 360PPI file that is a 16bit TIFF and (even better) is scanned pixel by pixel in some kind of drum scanner. But then someone will come along and win a contest with an 8 bit JPEG that a Fuji Frontier spat out at 8x12" resolution.

It should look right to your eye and your standards. That's about it! Knowing the most about optics, reproduction, and printing does not equal a great body of work (See: Ctein's portfolio). It goes a long way to know enough to be dangerous and stop there.

I'm sure this will be a post full of unpopular opinions.

 

[138S]

But then someone will come along and win a contest with an 8 bit JPEG that a Fuji Frontier spat out at 8x12" resolution.

Yes... but are you sure that the Frontier isn't cooking a 16bits/ch image before delivering a 8bits/ch result that has been treated with Image Intelligence?

8 bits are enough to print, and many images are perfectly described at 8bit depth, for example a portrait in the shadow with flat illumination.

Problem of 8 bits is that it does not allow to edit well extreme highlights and deep shadows, 8 bits depth has severe limitations for some images, and no limitation for other images.

The Frontier (or Noritsu) makes and amazing auto Pro edition that many are not able to emulate even if spending hours in Ps. Yes, no doubt, an straight Frontier image can perfectly win a contest because its Auto optimization is amazingly good. But if you want to control your image you may need a refined edition, and depending on the edition you need 16bits/ch to not make a botched job.

 

[grat]

Many scanners can do a multi-pass, it is a feature suported by SilverFast...

But in practice this won't be much worth.

Multi-exposure is to extend the dynamic range. It has nothing to do with detail vs. color.

Frequency separation is an image editing technique where the detail is separated out into one layer, and the color information is separated into another. This allows you to sharpen detail, without introducing artifacts into the color layer, or alternatively, to smooth out unwanted detail (think portrait editing) without "smearing" the color.

Ideally, you'd want a scanner with two sensors-- one B&W and one color. Scan at high resolution in the B&W scan to create the detail layer, then scan at a lower resolution for the color layer. It might be useless-- it might produce interesting results. I'll experiment, and if I find anything interesting, I'll pass it along.

 

[grat]

Yes... but are you sure that the Frontier isn't cooking a 16bits/ch image before delivering a 8bits/ch result that has been treated with Image Intelligence?

8 bits are enough to print, and many images are perfectly described at 8bit depth, for example a portrait in the shadow with flat illumination.

Problem of 8 bits is that it does not allow to edit well extreme highlights and deep shadows, 8 bits depth has severe limitations for some images, and no limitation for other images.

Define your terms-- 8 bits (total), with 256 colors, can produce some impressive images, but the palette has to be very carefully chosen.

8 bits per channel, for a 24 bit image, gives you 16 million colors to work with. Perhaps not the greatest tonal range, but still-- that's a lot of options.

 

[Lachlan Young]

As someone who scans 7 days a week for clients I can tell you I spend NO time thinking about MTF charts.

Once you're working with scanners above a certain quality level, it shouldn't (and doesn't) matter anyway! It's only the thread mangling defender of Epsons who apparently cannot see the blatantly visible difference with well maintained higher end machines/ camera set-ups.

Total contrast extintion at 6300 dpi effective at 1", which is contrast extintion at 1575 when zooming out x4 to catch 4" instead 1 with the same lens and sensor. Our tests delivered an exact inverse proportional yield to the scan width, which at all is not a surprise. Is it a surprise for you?

It's a prima facie indicator that you've never used an Imacon/ Hasselbad nor seen the preview screen in Flexcolor. 6300ppi is delivered across a 32mm width. And a contact copy (for this sort of purpose, the bare minimum would have to be vacuum platen, point source, developed to max contrast with appropriate developer) of a high contrast resolution target runs major risks of inaccuracy of contrast and resolution - given that you are trying to make claims on the basis of the resolution difference between two elements, the quality of your copy image is open to question at a minimum.

But then again, you must ask yourself why you and you alone think the Epson V700 is worth defending against much more expensive & higher performing scanners in the face of voluminous evidence to the contrary.

 

[warden]

As someone who scans 7 days a week for clients I can tell you I spend NO time thinking about MTF charts.

So many people get lost in techno jargon. Just set a target scan size that you need to achieve, and base that probably on the largest print size you're likely to make. When you make a scan, it should be roughly as detailed as you you expect it to be, with however much grain you also expect to see depending on the film. You can tell when your scanner is not up to the task (Epson) and when you may want to even doing a slight softening (Coolscan).

Lots of people will tell you that you need to have a 360PPI file that is a 16bit TIFF and (even better) is scanned pixel by pixel in some kind of drum scanner. But then someone will come along and win a contest with an 8 bit JPEG that a Fuji Frontier spat out at 8x12" resolution.

It should look right to your eye and your standards. That's about it! Knowing the most about optics, reproduction, and printing does not equal a great body of work (See: Ctein's portfolio). It goes a long way to know enough to be dangerous and stop there.

I'm sure this will be a post full of unpopular opinions.

Agree with all of that. Once you experiment with size and quality for yourself you don't need to invest in encyclopedic knowledge of scanning unless that's your hobby. (And there's nothing wrong with that as a hobby.)

My struggle is making anything worth printing (very hard) as opposed to making a fine print (pretty easy with the right help).

 

[flavio81]

A slight correction... well actually a fairly big correction to what you wrote. The Nyquist limit to resolve 125 lines per mm is 6450 ppi, which is double the figure you quoted. That is a lower limit to the sampling rate required, and as a practical matter a substantially higher rate would be required.

By the way, there is a subtle and seldom appreciated point with regard to the sampling theorem. The theorem says that if the sampling rate is at least as high as the Nyquist limit then it is possible to reconstruct the original sample without error. The theorem does not say that the sampled result is, in and of itself, an accurate representation of the original signal. To reconstruct the signal actually requires interpolating the raw data with a certain interpolating function to give more points. The correct interpolation function will produce a continuous function, but to get a good representation of the original function at discrete points requires resampling the reconstructed continuous function at discrete points, and the density of sampling points for the resampled result needs to be much higher than the Nyquist sampling rate. Otherwise, signals that are close to the Nyquist limit (but within the Nyquist limit) will be severely distorted. The correct interpolating function is based on the sinc function.

Claude Shannon is smiling on you from the heavens afar, knowing you have good knowledge of information theory.

 

[Lachlan Young]

making a fine print (pretty easy with the right help).

And this is the bit people want to make as seemingly difficult/ fetishistic as possible - to attempt to distract from the lack of imagery with anything to say...

 

[138S]

8 bits per channel, for a 24 bit image, gives you 16 million colors to work with. Perhaps not the greatest tonal range, but still-- that's a lot of options.

Many subjects have a single hue and saturation, so all r-g-b levels are proportional, in that case then when you have a gradation in fact you play with 256 levels, problem comes (for example) when you need to expand in (say) the 0 to 4 levels trande to say 0 to 32 range, you may get perceptible banding.

 

[grat]

Many subjects have a single hue and saturation

I'm pretty sure you're describing a monochromatic image. If you want to say similar, I'll agree that individual subjects (within a single photo) have a similar hue and saturation, sure. But that makes the argument *for* 8 bits per channel being sufficient, not against it.

Edit: Corrected poor phrasing

 

[Alan Johnson]

So to make a long story into a short question, if Fuji rates their film's resolving power at 125 lp/mm, which apparently translates to 3175 ppi, what's the benefit of scanning at a higher resolution?

With the Plustek 8100 scanning at resolution 7200 is reputed to give about 3600 lppm true and 7200 is a preset option in the menu ,easy to use so no point in not using it.
Even then the max resolution reached is less than that of the film:
https://www.photrio.com/forum/threads/resolution-of-the-plustek-8100.156680/
Have you not made an error in calculating, surely 125 lppm = 125x2x 25.4 =6350 ppi?

 

[138S]

It's a prima facie indicator that you've never used an Imacon/ Hasselbad nor seen the preview screen in Flexcolor. 6300ppi is delivered across a 32mm width.

You are right, I was mixing concepts of the X5 with the X1 that cannot scan 24mm width, so we just need to upgrade those numbers by a 6900/6300 factor, this is a 9% higher values. My mistake... please understand that I made that rating several years ago and I missed a 9% capability

So the right numbers are: 1725dpi at extintion for 4x5", which is total contrast extintion at poor 34.5cy/mm, so MTF 50% is obtained at 17.25cy/mm instead 15.5. Well, not nice for such an expensive scanner, quite good for 35mm but with a really limited capability for 4x5". At 1750dpi you have zero contrast !

But then again, you must ask yourself why you and you alone think the Epson V700 is worth defending against much more expensive & higher performing scanners in the face of voluminous evidence to the contrary.

This deserves a long explanation:

Because we paid quite a lot for Flextight and drum scans and later we realized that we could obtain mostly or exactly the same at home with the cheap Epson, we only had to learn how to use the Epson proficiently, by flattening well film and ajusting well height for perfect focus.

Later I had seen many V700 vs expensive gear side by side tests in what the V700 was machinegunned with forged results, comming from a poor handling:

>scanning auto with clipped highlights,
>no whitebalance of artificial light
>infame color edition
>no film flattening
>no focus fine adjustment
>no suitable sharpening
>no multiexposure usage when it was required

There is no doubt that a Creo or an Scanmate surpases the Epson technically by far. But also there is no doubt that the Epson is able to take all what a Portra shot may have, the reason is evident, in shooting conditions Portra resolves less than the Epson, let me reiterate:

In the vast majority of practical and real shooting conditions the lens is to resolve well less than 140cy/mm at extintion of the scene detail, and hence (with Portra 160) the image quality will be quite poor at only 40 or 50 cy/mm inspection, the real image is well at contrast extintion at those frequencies. BW usually does not go much farther in real shooting... with a flat target you may reach 90cy/mm with TMX, true, but real photography is made of 3D scenes and usually nothing is in perfect focus, or you may want to shot handheld... add that if something is recorded at 60 or 90cy/mm then it's is of very low quality...

The results in that 2019 comparison were not a surprise to me, I was smiling when that comparison was launched, I knew what the ratings were to be, but I also was knowing in advance the real/practical system yield and that the practical side by side was to be quite equal. For the first time an honest side by side was performed, and also I was suspecting that the V700 was to be operated proficiently, as it was.

 

[Adrian Bacon]

I believe you have shot right past my point.

I'm proposing a different methodology for scanning a color negative to produce an image format. What I got from the article was that the grain resolution is far higher than the color resolution, suggesting a two-pass approach to the scanning process.

I got your point. You’ll get results similar to how images are currently stored using the same methodology, except perhaps a bit more resolution due to the higher resolution black and white scan. It’ll let you get a bit more luminance resolution, which is what we’re most sensitive to.

 

[138S]

I'm pretty sure you're describing a monochromatic image. If you want to say similar, I'll agree that individual subjects (within a single photo) have a similar hue and saturation, sure. But that makes the argument *for* 8 bits per channel being sufficient, not against it.

Edit: Corrected poor phrasing

for example you have the sky, or water bodies, or buildings... if you don't manipulate much the curve you won't have problems in 8 bits, but there is no doubt that high contrast scenes needding a deep edition require 16bits

Also... sharpening algorithms are to work much worse in the deep shadows, if deep shadows are enconded 1 to 16 levels, for example, operations are truncated delivering way less precision, if later you expand those shadows.

A proficient edition is in 16bits, still 8bits can be totally acceptable in many situations, this is also true.

 

[Helge]

Claude Shannon is smiling on you from the heavens afar, knowing you have good knowledge of information theory.

I think he was smiling at me first. ;-)