Measuring film resolution

Nodda Duma · May 24, 2015

Look up slant edge MTF testing. There are a couple of software packages for < $100. The only target required is a high-contrast sharp edge set at a 10 degree angle. I've used the technique professionally for years testing military imaging systems when using an optical MTF station was not feasible.

If you don't change anything except the variable under test (film/developer), then you should be able to qualitatively compare performance if not strictly performing a quantitative comparison.

Rudeofus · May 24, 2015

Nodda Duma: I am aware of slant edge tests, but they require a scanner with much higher resolution than your film. My hope was that a resolution target would avoid this requirement.

@RobC&Old-N-Feeble: Some years ago I shot Delta 3200 in 35mm format, which I developed in HC-110 to EI6400. Portraits where the face filled more than 20% of the frame worked fine, but shots with smaller faces were unusable due to grain. "Worked fine" in this context means "looks good if enlarged to 13x18cm", not mural sizes. This was a case where excessive grain all but made unusable a whole roll of film, and no tripod or Zeiss lens would have made it usable.

Alan Johnson · May 24, 2015

There are tables of film resolution like this:
http://www.cacreeks.com/films.htm
But they are not so useful if you actually want to see what the printed lines look like.

Nodda Duma · May 24, 2015

Rudeofus said:
Nodda Duma: I am aware of slant edge tests, but they require a scanner with much higher resolution than your film. My hope was that a resolution target would avoid this requirement.

The nice thing about slant edge test is that it specifically doesn't require scanning at resolution much higher than that of your imaging media (film in this case). Whatever you usually scan negatives at is probably sufficient.

Rudeofus · May 24, 2015

Nodda Duma said:
The nice thing about slant edge test is that it specifically doesn't require scanning at resolution much higher than that of your imaging media (film in this case).

This means I would need at least 5000 real dpi for measuring 100 lp/mm. Sorry, my V700 is way below that, and so are most but the most expensive scanners.

Nodda Duma · May 24, 2015

Well, don't try it then . But it's the least expensive method I can think of.

Athiril · May 24, 2015

Rudeofus said:
A contact printed resolution target tells you both limit resolution and which image property dominates that limit. It is entirely up to you whether you favour sharpness over grain, but you'd rather make an informed decision than rely on anecdotal evidence or marketing drivel.

Realistic shooting conditions won't give you more than 100 lp/mm. To give you an impression:

For 100 lp/mm your CoC can't be larger than 2-3 µm, which means you have about 1/10 of DoF which you would normally have, e.g. a 50mm lens at F/2.8 and 3m subject distance would have 4cm of total DoF). Good luck getting focus sufficiently accurate, and subject matter with less depth.

Because of diffraction limits, you have to shoot at F/2.8 or larger aperture if you want 100 lp/mm. Few medium format lenses open up to F/2.8, and even fewer large format lenses. Very few small format lenses reach 100 lp/mm at F/2.8 or larger.

Extend this argument to resolutions beyond 100 lp/mm and see what you get. 200 lp/mm would require you to shoot at F/1.4 and give you 1 centimeter of DoF in the example above.

Now if we look at the Zeiss document where they claim 400 lp/mm at F/4: we can safely assume that these 400 lp/mm come at a significant loss of contrast, which is ok with test charts, but not ok with real world image matter.

There are some tricks to increase perceived sharpness of an image, and several popular developer formulas put these to use. These tricks don't increase real resolution, but fake the impression of a very sharp image, just like unsharp masking. Like the latter, these developers usually create stronger granularity. And that's where Adox CMS 20 film becomes useful: even if you boost its grain a great deal, it's still finer grained than what your lens can handle, and together with a sharpness enhancing developer you can create very fine grained and very sharp looking images, albeit at ISO 20.

I can hit over 100 lp/mm. Using a regular camera, holding the camera down on a flat surface with my hand, in 6500k+ light.

I've modified commercial developer formula to give both a real improvement in both grain and resolution, and an increase in resolution gives an increase in sharpness/IQ at the same frequency over two samples, as it is more clear. It's a real increase in resolution.

Also an increase in resolution, is an increase in high frequency contrast, not a loss, hence if you compare a max of 50 lp/mm and a 100 lp/mm, the 100 lp/mm sample, will bu superior @ 50 lp/mm, and still look good beyond that.

Paul Verizzo · May 24, 2015

After a half century of trying to find perfection, my new mantra is, "Take pictures."

How many of the "greats" sweated these details? None.

piu58 · May 25, 2015

> After a half century of trying to find perfection

"Have no fear of perfection - you'll never reach it." - Salvador Dalí

Athiril · May 25, 2015

Rudeofus said:
I have seen many characteristic curves for film&developer combos posted here, but for full evaluation of film&developer we also need to look at resolution. For this purpose I would contact print resolution targets onto my film and then try to extract the relevant data from developed test clips.

Looking for suitable contact printing masks gave me two categories of product: simple gel targets like the one offered by Stouffer, which don't cover the range of resolutions achieved by modern film, and very expensive very high resolution glass targets like the ones made by Edmund Scientific, Thorlabs and others.

Is there something in between, which covers resolutions up to 100-150 lp/mm while costing less than US$100? How do others here measure film resolution? Are there affordable targets for measuring MTF from 20-100 cycles/mm? Is there a dedicated used market for such things?

To expand on this I printed my own chart, max contrast was about 75:1 on mine, made a variable contrast one as well. Puts 150 lp/mm+ without much effort on CMS 20, scanned (high end scanner) and inspecting with inexpensive microscope.

If you want to make something on the cheap, you might want to look into multiple slit diffraction interference (like the double slit experiment), there's equations and calculations for calculating the approximate width of the line you get (or frequency/resolution), based on slit distance separation and distance to screen.

I've thought about it quite a bit before, If I remember there's a number of slits that would be more ideal for this type of thing then simply double slit, and if you made slits that curved away from each other line a line chart, the frequency/resolution would increase with the increasing separation and you'd have a curved line chart you could print directly..

I think the line width also changes from center line to outer lines if I remember, you probably can counter act that by curving with a radius from the center of the film to the center slit.

Rudeofus · May 25, 2015

Athiril said:
Also an increase in resolution, is an increase in high frequency contrast, not a loss, hence if you compare a max of 50 lp/mm and a 100 lp/mm, the 100 lp/mm sample, will bu superior @ 50 lp/mm, and still look good beyond that.

Yes, some developers give better overall resolution than others, but then there are developers which trade apparent sharpness for real resolution. A typical example for this is unsharp masking, which can be done by creating Mackie lines during development, by suitable masking during enlargement, or in digital post processing.

Paul Verizzo said:
How many of the "greats" sweated these details? None.

Quite a few of them, some of them devised their own developer formulas, others wrote several renowned books about darkroom technique.

If image quality was of no concern to folks here, why do they go through all the hassle that comes with analog photography?

L Gebhardt · May 25, 2015

Lasersoft imaging sells a USAF 1951 target for $60 (35mm slide mount and not glass). It has group 7, so in theory it has up to 228 lp/mm. If you remove it from the slide mount and contact print it under even pressure it might work for your needs.

http://www.silverfast.com/show/resolution-target/en.html

Paul Verizzo · May 25, 2015

Rudeofus said:
Yes, some developers give better overall resolution than others, but then there are developers which trade apparent sharpness for real resolution. A typical example for this is unsharp masking, which can be done by creating Mackie lines during development, by suitable masking during enlargement, or in digital post processing.

Quite a few of them, some of them devised their own developer formulas, others wrote several renowned books about darkroom technique.

If image quality was of no concern to folks here, why do they go through all the hassle that comes with analog photography?

The thread isn't about "image quality," but resolution, mostly lens, sometimes developer. Only one aspect of image quality. When lenses were very difficult to calculate and manufacture, how where they evaluated, especially at the user level? I'm guessing no expensive AF charts in Matthew Brady's kit. Probably take a picture of a newspaper and examine and decided "Good," or maybe "Not so good."

RobC · May 25, 2015

I think, although he hasn't speciffically said, the topic is about being able to test the effect of different developers on resolution and so has nothing to do with lenses.

Rudeofus · May 25, 2015

If film resolution was mostly an emulsion characteristic, why would people work with PPD based ultra fine grain developers, which lose two stops to go? Obviously you can't turn TMAX 400 into TMAX 100 with these developers, but I would expect a visible difference in image structure between TMAX 400 developed in Dektol vs. Microdol-X.

From a signals processing perspective, edge tests are very inaccurate unless the analysed medium is mostly noise free. People tried to avoid readout noise by averaging over very long edges, but the MTF method is clearly better. Let me provide you some details:

In the end, all these methods try to determine the resolution of a medium. Our medium suffers from two issues which both reduce resolution: noise (aka grain) and blur (aka lack of sharpness). Adding blur will reduce sharpness&grain, and correctly applied unsharp masking will increase sharpness&grain, so you can freely switch between these two for a given medium.
Resolution is usually measured as something per millimeter, i.e. it is a measure in frequency space. Since noise also goes up linearly with inverse aperture diameter, even a perfectly sharp but noisy medium will have a resolution limit.
Resolution doesn't fall off a cliff when one approaches a medium's limits.Depending on noisiness and sharpness you either lose contrast to the point where you can't reliably distinguish input contrast, or image noise becomes so extreme that image matter is indistinguishable from noise patterns.
Measuring noise is fairly easy: scan in an area that received uniform exposure and development, and measure the variance of density. Since the area one measures this variance in can be large, reliable numbers can be obtained easily.
Measuring the frequency response of a medium can be done by differentiating the step response of that medium, or by measuring its frequency response directly. The first method is susceptible to numerical inaccuracies, since numerical differentiation amplifies noise, especially high frequency noise. Frequency response measurement for each data point can be done over a large area (in two dimensions, not one!) and can therefore trivially cancel out noise. The second method is known as MTF measurement, and became the predominant method once people understood the mathematics behind it.

@Paul Verizzo: before any lens reaches a consumer today, its design has gone through extensive simulation and evaluation by the maker, including but not limited to advanced resolution tests. These hands on tests posted by various online publications only state things already known to the lens' manufacturer, and surprises happen only if marketing tries to sell something that engineering couldn't deliver.

@L. Gebhart: thanks for the link! I looked at a similar offering by filmscanner.info, but their excessive shipping costs put their target into the same price range as the professional resolution targets. Lasersoft's total price seems a lot more reasonable.

Conclusion: at the moment Lasersoft's target, as suggested by L. Gebhart, looks like the way to go. I still wonder whether there is a used market for such professional targets. Twenty years ago, there were several places for used lab equipment at discount prices in the Silicon Valley alone.

Old-N-Feeble · May 25, 2015

Look online for comparisons between high-resolution high-acutance developers vs. low-grain solvent-type developers. At the extremes there is a very pronounced difference in sharpness. Choose wisely depending on subject matter and desired final look of the image. I would never use a solvent-type developer to test film sharpness.

Athiril · May 25, 2015

I've found Rodinal 1+25 and Xtol Replenished to have very close resolution.. a least on FP4+.. near doubling the linear resolution (squaring the detail) with finer grain in both with the addition of potassium thiocyanate and potassium iodide. I also found that Rodinal 1+100 stand with FP4+ reduced resolution.

Here is some T-Max 100 (image is 200% digitally enlarged to see the chart a bit easier), Rodinal 1+25 top, Rodinal 1+25 + additions at bottom.

Photo Engineer · May 25, 2015

Guys, a film can be built like shown in the figure below. It will have higher contrast as the size of the image decreases or the size of objects in a constant image size decreases. It can also be built such that the micro and macro contrast are the same. This puts all of your tests through hoops and gains nothing for you unless you shoot and print, thus seeing results from the entire image train.

OTOH, we have been through this so many times, you might want to read, or re-read, such as the case may be, the article by Tim Vitale. This article was posted as a URL long ago on APUG, but I downloaded it to read at my leisure. It is, sadly, too big to upload. I may try zipping it but I'm rather busy right now so I'll let you guys check it out.

PE

albada · May 25, 2015

Athiril said:
I've found Rodinal 1+25 and Xtol Replenished to have very close resolution.. a least on FP4+.. near doubling the linear resolution (squaring the detail) with finer grain in both with the addition of potassium thiocyanate and potassium iodide. I also found that Rodinal 1+100 stand with FP4+ reduced resolution.

Here is some T-Max 100 (image is 200% digitally enlarged to see the chart a bit easier), Rodinal 1+25 top, Rodinal 1+25 + additions at bottom.

Athiril, that is a big difference! Can you post the formula (additions) you used for the second image? Also, how much film-speed was lost? Do you have the same comparison-images using XTOL?

Thanks,

Mark Overton

Athiril · May 25, 2015

albada said:
Athiril, that is a big difference! Can you post the formula (additions) you used for the second image? Also, how much film-speed was lost? Do you have the same comparison-images using XTOL?

Thanks,

Mark Overton

I didn't notice any, but you may have to do your own tests I haven't tested for shadow detail loss.

iirc it's 1.4g/L of potassium thiocyanate and 10mg/L of potassium iodide added to working developer, use as normal.

I've got Xtol ref comparison and also added to Xtol, some films build up dichroic fog this way in Xtol (none of the films do this in Rodinal), but it can be wiped off and still shows an improvement, the T-Max improvement is better in Rodinal + ingredients then Xtol, works on most of the films I've tested, I didn't get a resolution increase on Delta 3200, only grain improvement, but haven't tested that enough.

I'll post the reference examples and others later.

Old-N-Feeble · May 25, 2015

Athiril... I swore I wasn't going to get into chemical modifications anymore but your post looks darned promising.

Athiril · May 25, 2015

It wont improve flatbed scanning results, but anything that is reasonably high resolving.. good scanners, and enlarging with reasonable lenses, with the original image taken with a reasonable lens, then you should see results, I haven't tested all films. Next I want to test is Delta 100 and Acros 100.

Both these are available inexpensively on eBay worldwide (most countries) if needed, my KSCN was tested (assuming for drugs) by customs.

Athiril · May 25, 2015

albada said:
Athiril, that is a big difference! Can you post the formula (additions) you used for the second image? Also, how much film-speed was lost? Do you have the same comparison-images using XTOL?

Thanks,

Mark Overton

Xtol Replenished seems a bit better than Rodinal 1+25 for T-Max for me, Xtol Replenished with the additions a bit better again, and Rodinal 1+25 with the additions being the best.

Digitally enlarged 200% (nearest neighbour) to make peeping easier.

Might need to click and open in a new tab/window to see full size

Photo Engineer · May 25, 2015

Dan, do you have H&D curves to go with these 4 samples?

PE

Athiril · May 25, 2015

No, it's not what I was looking for in this case. I have access to densitometer, but not a sensitometer. Unless someone wants to donate one, or send two strips of T-Max 100 that've had a scale exposed on them to me to process the best I could do is metering something as 18% grey through a camera, exposing it up and down the scale and plotting the results vs exposure.

Measuring film resolution

Attachments

Privacy & Transparency

Privacy & Transparency