PDA

View Full Version : Scanning Resolution



robertrose
27-Mar-2015, 17:30
There are several threads related to specific scanners; here I just wanted to suggest that worry about super-high resolutions such as 6400 dpi for 4x5 film is overkill, and will result in many unhappy hours at the computer.

For example, suppose that you are going to print a 16x20 print on the popular Epson 3800 or 3880. You want at the most 360 dpi for the output file. This means a 5760x7200 image for a 360 dpi print. To get that from a 4x5 negative you only need 1440 dpi. Even at that lower resolution you will downsample for web size images, or if the subject matter requires (see http://robertrose.photos/galleries/large-format/yosemite-river-2008.html).

There are many other issues if you are printing, such as whether to use 8-bit or 16-bit files, etc., and I refer you to Eric Chan's excellent website http://people.csail.mit.edu/ericchan/

There is an excellent set of articles about scanners at the LF home page, but in all honesty I have yet to see a single person examine my prints with a magnifying glass. It is subject matter and composition that makes the photo.

Peter De Smidt
27-Mar-2015, 17:40
<snip>
For example, suppose that you are going to print a 16x20 print on the popular Epson 3800 or 3880. You want at the most 360 dpi for the output file. <snip>

What is your evidence for this?

Jim Jones
27-Mar-2015, 19:33
Scanning at exactly the right dpi to print the largest print I expect to make seems logical, but scanning at higher resolution and carefully editing that file allows for perhaps future printing at even larger sizes. I agree with Robert that any downsizing and final editing should be done in the computer at the printer's native resolution. It works for me.

Preston
27-Mar-2015, 20:22
Scanning at exactly the right dpi to print the largest print I expect to make seems logical, but scanning at higher resolution and carefully editing that file allows for perhaps future printing at even larger sizes.

Indeed, it does seem logical, and would certainly work well for a person who does not have a computer with the horsepower and/or storage for huge files. Personally, I scan at 1800 ppi (the max for my Microtek 1800f and Vuescan) to produce a 16 bit TIF file. This file is then saved as a PSD 'master file'. All my editing is done on this file. Output files are for printing or the Web.


I agree with Robert that any downsizing and final editing should be done in the computer at the printer's native resolution.

Same here. If I need to make small adjustments to a print file or Web file, I do exactly as Robert and Jim suggest. Works for me, too.

--P

vinny
28-Mar-2015, 06:47
So..........if i want to make a 16x20 I should scan my film at 1440dpi, open the file in photoshop, spend a considerable amount of time removing dust, make any adjustments, then finally print it.
Then when I get an order for a 32x40, I should fire up the drum scanner, remount the film, and do it all over again?
Thanks for the suggestion but, NO THANKS.

Bill Burk
28-Mar-2015, 07:49
Don't you work on a managable proxy file, save the actions and original hi-res, then anytime you get an order... downscale copy of original to optimum for target output print and then apply the saved actions to it?

Peter De Smidt
28-Mar-2015, 07:54
I was under the impression that the native resolution of some Epson printers, with "finest detail" checked, is 720 dpi.

See Eric Chan's comments on print quality at: http://people.csail.mit.edu/ericchan/dp/Epson3800/faq.html#native_res

paulr
28-Mar-2015, 08:10
"Finest detail" is a raster setting optimized for line art; it's not ideal for images with photographic gradients.

There are other setting that give 720ppi that are appropriate for photographs. They're only available with gloss paper presets.

I'd recommend doing a blind test (blind as in, you don't know which is which), preferably with three prints of an identical image. In my experience it's usually impossible for anyone to tell the difference between 360 and 720ppi. The only exceptions I've encountered have been with crisp diagonal lines that are just a bit off from vertical or horizontal. Sometimes at 360ppi you'll see some slight aliasing (jaggedness of the line).

I don't believe with a photographic image, anyone will see more detail or subjective sharpness at 720ppi, all else being equal. When I was printing a body of black and white work from 4x5 scans using piezography, I did it at 720, because it felt badass. But I didn't have a rational reason to do so.

djdister
28-Mar-2015, 08:14
This is a losing argument. Been there, done that. Optimum scanning resolution should be maximized for whatever you intend to do with the final image file. Its as simple as that.

paulr
28-Mar-2015, 08:16
My approach to scanning has been to run tests to find the highest resolution that makes a visible difference. With my old 4870, with shims and a wet-mount setup, I could see the difference between 2000ppi and 1800 (barely). I could not see the difference between 3600 and 2400.

I settled on scanning at 2400, because it divides evenly from the scanner's sampling frequency, and allows plenty of downsampling to help control noise.

Years ago I would immediately downsample to print resolution, since drive space was precious and the computer sluggish. I also often worked on a half-resolution proxy file. but today I'd probably work on a full-resolution master file and then make size-specific versions for printing.

Some people use a program like Vuescan and just save the raw scan file. It's big, but it grabs every bit of information the scanner produces (resolution, gradient, infra-red noise channel, etc..) so you can go back to this file and reprocess at any time you like.

Peter De Smidt
28-Mar-2015, 08:21
Thanks for the info, Paul. I use QTR not the Epson driver, and so I'm not very familiar with the latter.

To the original poster, for prints less than 30", it's not hard to have a scan from 4x5 that'll give 720 dpi at output size. Today, digital storage is cheap and many computers can easily handle the big images. Why throw resolution away if you have it? Why use interpolation if you don't have to?

paulr
28-Mar-2015, 08:25
Optimum scanning resolution should be maximized for whatever you intend to do with the final image file. Its as simple as that.

Yeah, that's one approach, but it presumes that you either know ahead of time the largest print you'll ever make, or that you're fine with scanning the same neg again if the need comes up. Personally, I'd like to avoid rescanning as much as possible.

paulr
28-Mar-2015, 09:17
I never suggested rescanning your neg. That was your presumption.

You're suggesting scanning for the biggest print you intend to make. This means either 1) you're positive you're not going to make a bigger one at some point, or 2) you don't mind rescanning.

Peter De Smidt
28-Mar-2015, 10:00
Dan, why not explain yourself instead of being rude?

Your statement: "Optimum scanning resolution should be maximized for whatever you intend to do with the final image file. Its as simple as that."
could mean: A) Scan to retrieve the most resolution that your scanner is capable of, or B) scan so that your print couldn't be improved by different scanner settings at the sizes you intend to print at.
It looks like Paul interpreted what you said as B, whereas you intended A.

Ken Lee
28-Mar-2015, 10:42
Rude posters get banned: briefly at first, then permanently when necessary.

Lenny Eiger
28-Mar-2015, 19:13
Don't you work on a managable proxy file, save the actions and original hi-res, then anytime you get an order... downscale copy of original to optimum for target output print and then apply the saved actions to it?

Nope. I have a computer that can handle what I am doing. I like just doing it right the first time. Nothing takes that long anyway...

I will also add the concept of "archiving" the image using the scanner, doing the scan to as much of the original image as is technologically possible, as in drum scan. Then if something happens to the film you still have the scan.

Lenny

onnect17
28-Mar-2015, 19:39
AFAIK the rendering limit for the epson drivers (at least in the old Windows systems and printers) is set by the 16 bit integer (32767). For example, if I try to print a 40"x40", then 32767/40 = 819, so the closest is 720dpi. If I try to print a 40"x60" then 32767/60 = 546, bringing the limit number to 360dpi, and so on. Pretty much any dimension over 100" and the rendering will be done at 180dpi.

The point is, no matter how big your original image is, Epson windows driver will only be able to handle a page with less than 32767 pixels per dimension.

There's a printing application for Windows I am sure many here are also using called Qimage. This app gets the most of the Epson driver and the interpolation algorithms are quite decent. Worth every penny.

paulr
28-Mar-2015, 21:37
The point is, no matter how big your original image is, Epson windows driver will only be able to handle a page with less than 32767 pixels per dimension.

That's interesting. Where did you find this information?

32,000 is still a hell of a lot of pixels. This would only of concern to someone printing murals from 8x10 or bigger.

Kirk Gittings
28-Mar-2015, 21:41
but in all honesty I have yet to see a single person examine my prints with a magnifying glass. It is subject matter and composition that makes the photo.

Then why shoot large format at all? You could shoot 6x7 on a small VC or use tilt shift lenses on a DSLR?

onnect17
28-Mar-2015, 22:42
That's interesting. Where did you find this information?

32,000 is still a hell of a lot of pixels. This would only of concern to someone printing murals from 8x10 or bigger.

Search for Epson support, Article ID:3249
"What's the maximum image length allowed when printing a banner or long image on roll paper?"
...
When printing using an Epson large format printer and the standard Epson driver the length of the image cannot exceed 32767 pixels.
...

robertrose
28-Mar-2015, 23:06
Fair question. Instructions from Epson; my own judgment when printing on the 3800 to Epson paper; and Eric Chan's website, a Principal Scientist at Adobe:
http://people.csail.mit.edu/ericchan/dp/Epson3800/faq.html#native_res
"If the 'Finest Detail' driver setting is unchecked, the 3800's native resolution is 360 pixels per inch (ppi). If the 'Finest Detail' driver setting is checked, the 3800's native resolution is 720 ppi. For most photographs, I recommend unchecking (disabling) the 'Finest Detail' driver setting and preparing your final images at 360 ppi."
Chan explains that the 720 setting is for vector-based data including text, graphics, and line art. He speculates that prints will not benefit from higher resolution higher than 360, and that accords with my experience.
I have no doubt that your experience might differ, especially with different papers and inks, and recommend experiments.

robertrose
28-Mar-2015, 23:17
Yes, but he notes that the higher 720 dpi makes a difference mostly for line art. He also makes the point that your results may differ, and I certainly agree with that.

robertrose
28-Mar-2015, 23:21
I salute you; I don't have a way to print 32x40 and no one has ordered that big a print from me. Yet.

Jac@stafford.net
29-Mar-2015, 08:17
When printing using an Epson large format printer and the standard Epson driver the length of the image cannot exceed 32767 pixels.
...

I haven's used that unsigned 16bit int since programming PDP-11s. Perhaps someone smarter than I can explain why Epson is limited to such. Architecture? Little old chips?

Ken Lee
29-Mar-2015, 08:19
I found no tangible benefits printing at 720 dpi - only inconvenience. I went back to 360 dpi. I like it better.

lbenac
29-Mar-2015, 08:30
Excuse by advance my lack of knowledge but I seem to recall reading instructions for Piezography ink printing on QTR to the effect that it was "better" to send the image to QTR at full pixel strength (no down-sampling to 360 dpi) and let QTR do that portion.
QTR I understand that QTR doe snot use the Epson driver so it might explain the difference?

Cheers,

Luc

paulr
29-Mar-2015, 08:37
"Finest detail" is a raster setting optimized for line art; it's not ideal for images with photographic gradients.

I'm retracting this; it's what I was taught, but I think Peter is right; this just sets the driver to 720ppi native resolution. I don't believe there's any difference that would be detrimental to gradients.

I haven't tested this setting on or off; it would be worth a few experiments. In case it isn't obvious, it only makes sense if the resolution of your file is higher than 360ppi without any upsampling.

Peter De Smidt
29-Mar-2015, 10:02
I found no tangible benefits printing at 720 dpi - only inconvenience. I went back to 360 dpi. I like it better.

Printing on matte paper? I expect that there's only a noticeable difference, if there is any at all, with glossy paper.

JChrome
29-Mar-2015, 11:19
Interestingly enough, not only do you not need higher resolution but you might also prefer a "lower" quality scanner.

I took the below photo during 2014's burning man. A friend showed great interest in it and so I decided to give him a print. The below was scanned using a canon 9000f flatbed (something like $150.00). The negative is a 6x9 portra 160. Notice the dust marks (this was shot during a dust storm and it was impossible to keep that fine dust from the film).

Wanting to get the "best of the best" I drum scanned at 5000 dpi and got a beautiful 2.5 gigabyte tiff file. Wet mounting the scan on the drum scanner smoothed out the dust marks sufficiently so that they are almost* imperceptible.

The flatbed's light refracted the dust marks on the film and exaggerated them. But I think this is what my friend wants! The dust spots starts a conversation that just isn't there without the dust.

What do you gents think?

Flatbed Scanner (with dust)
131570

(Drum scanned version to come)

fishbulb
29-Mar-2015, 11:33
This conversation reminds me a lot of the scene in Star Wars Episode III:


Anakin Skywalker: If you're not with me, then you're my enemy.
Obi-Wan Kenobi: Only a Sith deals in absolutes.
[draws his lightsaber]

Everyone has a different workflow.

For some, scanning is labor-intensive, so they scan once at the highest possible/useful resolution.
For others, they don't mind rescanning if they want to print larger..
And others even prefer dust spots and scanning artifacts, apparently.

Each to his own.

The Deer Gunter
30-Mar-2015, 04:23
It surprises me so far no one talked about the resolving power of film, and more in particular the different film formats. It's not only a question of how thick the emulsions are and what film speed you're using, but also and mainly how much detail photo optics CAN deliver to the photographic emulsion. The bigger the format, the more information you get from the emulsion for sure, but the overall sharpness goes down due to a typical bigger circle of confusion.
In this respect I think it's always wise to use the highest optical resolution your scanner is capable of when scanning small format (35 mm film), but a 4x5 inch would have very little benefit, if no benefit at all, from a 5 000 ppi scan. The only thing you would gain, is weight, but no true image detail. Even more, I think oversized images are more difficult to sharpen, as the image's grain acutance becomes more fuzzy, causing the transitional contrasts harder to be detected and isolated.

Do you agree? And what do you think a reasonable scan resolution would be for a 4x5 inch slide?

mdarnton
30-Mar-2015, 04:43
Do you agree?
I don't think I do. Last spring I got my first 8x10 camera, and set it up pointing out my shop window. Twelve floors up, I can see several blocks. Putting the loupe to the GG I was very surprised at the detail I could see--with a better loupe it would have been very easy to read license plates a couple of blocks away. Then I had what was for me a big revelation: I was using a 300mm Tessar, a very sharp lens that resembled many 35mm teles, and looking at a 35mm size piece of film. Of course, that would not be an unusual situation for 35mm, and think of the detail you would naturally expect in that situation. There's no reason at all not to expect it in the middle of a larger piece of similar film behind a similar lens, right?

So if everything is well-aligned, I would expect if you can scan 35mm and get useful detail at 5000ppi, you would get useful detail from a similar film/lens combination at 5000ppi if the film happened to go farther out in every direction (i.e., 8x10). Whether you can use it making an 11x14 print is, of course, a different question. But it's there to use.

The Deer Gunter
30-Mar-2015, 06:57
There's no reason at all not to expect it in the middle of a larger piece of similar film behind a similar lens, right?


I'm afraid there is! If you would compare a true 8x10 inch lens (covering the 8x10 inch usable image circle) to a true 35 mm 'tele' lens giving you the same enlargement on the 35 mm film, you would see a significant difference in sharpness! And I'm not talking about whether you see that license plate on the ground glass or not. Viewing your ground glass with a magnifier is not a reference at that point. I'm talking about the acutance of the details of both shots when you scan at high res, or even when you view them with a 25x loupe on the light table. You would be surprised to see the differences between both recordings! 35 mm Lenses reach a higher degree of sharpness. The Circle of Confusion (CoC) for 35 mm lies typically around 0.029 mm, whereas an 8x10 inch format system would reach a CoC of 'only' 0.22 mm!

This is the reason why I think you would be overkilling your large format scan at extreme high scan resolutions, because the resolving power is simply not available.

onnect17
30-Mar-2015, 07:20
IMHO, 5000 dpi is OK with 4x5 with the right lens, emulsion and scanner.
Nowadays with apo, ed, and very fine grain we can get near 100 lp/mm with that particular format.
Perhaps would be wise to verify the real resolution of the scanner in use.

Peter De Smidt
30-Mar-2015, 07:43
It's not just resolution on subject, though, that matters for prints. It's also how the grain is presented. For instance, I've scanned 35mm film at a wide variety of resolutions with my Cezanne. The best results, by a significant margin, came at 6000 spi. At that setting the grain was much finer and more even than at 4000 spi. Of course if you're not enlarging enough to see grain, then that probably doesn't matter.

For me, the balance between quality and cost goes 6000 spi for 35mm, 4000 spi for 120, 4000 spi for 4x5, 2500 spi for 8x10; but this is something that everyone who scans should check out with their film and their scanner.

Jim Jones
30-Mar-2015, 08:00
. . . The Circle of Confusion (CoC) for 35 mm lies typically around 0.029 mm, whereas an 8x10 inch format system would reach a CoC of 'only' 0.22 mm! . . .

The CoC is not an inherent quality of the lens; it is an arbitrary figure created to facilitate using the lens. Even though some lenses are designed for optimum sharpness in small formats, extensive lens tests done by others show some long LF lenses having fairly comparable sharpness over their larger film area.

The Deer Gunter
30-Mar-2015, 08:11
It's not just resolution on subject, though, that matters for prints. It's also how the grain is presented. For instance, I've scanned 35mm film at a wide variety of resolutions with my Cezanne. The best results, by a significant margin, came at 6000 spi. At that setting the grain was much finer and more even than at 4000 spi. Of course if you're not enlarging enough to see grain, then that probably doesn't matter.

For me, the balance between quality and cost goes 6000 spi for 35mm, 4000 spi for 120, 4000 spi for 4x5, 2500 spi for 8x10; but this is something that everyone who scans should check out with their film and their scanner.

I think you may be very close to reality Peter! The only thing I would estimate higher is the medium format resolution. I still found extra usable image information at 5 500+ spi when scanning Mamiya 7 shots. I never took it higher up, but I felt all the detail the film could offer was squeezed out of the lemon at that point.

fishbulb
30-Mar-2015, 08:17
...this is something that everyone who scans should check out with their film and their scanner.

I quite agree.


...you would be overkilling your large format scan at extreme high scan resolutions, because the resolving power is simply not available.

This is only true if the photo was taken very stopped down (f/32, f/64, etc.) At wider apertures, there is a great resolution benefit to scanning large format at 4000dpi, and even minor benefits for scanning at even higher resolutions.

With my own tests of 4x5 Delta 100, I saw significant increases in detail when scanning the same negative at 1000dpi to 2000dpi to 3000dpi to 4000dpi on my Howtek 4500 drum scanner. So I scan everything at 4000dpi and scale it down as needed. (I rarely stop down more than f/22).

This conclusion is supported in detail here: https://www.onlandscape.co.uk/2011/12/big-camera-comparison/ Check out the first data table. 4000dpi is about where you get about the maximum resolution on 4x5 and 8x10 film. 8000dpi adds a little more real resolution (line pairs) on finely-grained film like Delta 100, but not much more on other films.

Take a look at the second data table in that link. 4x5 and 8x10 suffer large drops in resolution as they are stopped down, so your hypothesis is correct IF the particular photos were taken at small apertures (large f/numbers). Which is may be likely - it's common to see large format images taken at f/32 or f/64. So 2000dpi is probably more than enough for those apertures.

However, when I see a photo taken at f/64, I wonder if the photographer could have used some movements to adjust the plane of focus and taken it at f/32 or even f/16, and really increased the resolution of their image.

A later update to the above linked article includes 35mm film (Canon A1) and 36mp digital (Nikon D800e). https://www.onlandscape.co.uk/2014/12/36-megapixels-vs-6x7-velvia/
And this one has more info on print sizes, how big you can print, etc.: https://www.onlandscape.co.uk/2015/03/resolution/

8x10 user
30-Mar-2015, 08:59
360 PPI from a 2880 DPI print engine (screen) is quite nice. I've had reports from major suppliers of RA4 imagers that the actual resolution that they are obtaining is closer to 160 PPI. The samples I got from the US based RA4 imaging supplier sucked when compared to the newest inkjet technology. There is a European based equipment manufacture that makes an RA4 machine that is even higher quality, IMO, then todays inkjet.

In terms of scanning resolution I feel that if you have a good scanner that it is worth it to scan at the highest resolution possible. Even if you reduce the resolution later it will give you better tonality and will capture fine details and film grain with more accuracy. If you capture the grain better then it is easier to remove. Ideally your sample size should match the average grain clump size but the rate of sampling should be up to 3 times higher (due to the stochastic nature of film).

paulr
30-Mar-2015, 09:37
Ideally your sample size should match the average grain clump size but the rate of sampling should be up to 3 times higher (due to the stochastic nature of film).

I'm assuming you're talking about drum scanning. With flatbed scanners, the maximum sampling frequency is usually much higher than the optical resolution. This gives automatic anti-aliassing, and is good for marketing. But it also means that the only real utility to oversampling is noise reduction.

Oren Grad
30-Mar-2015, 10:04
I haven't tested this setting on or off; it would be worth a few experiments.

I've done the comparison, many times. I can pretty consistently tell the difference between 360 and 720, on every paper I've used.

BUT, and it's a big but: the reason this matters to me is that I'm often printing big files small, with as much subtlety as I can muster, with the intent that the prints stand up to a ridiculous level of scrutiny without conveying any impression of being "processed". For example, the files from my 6D are 5472 x 3648 pixels. When I'm in the mood to do this kind of printing, I'll send 6D files to the printer at 720 ppi with no sharpening or interpolation of any kind, to make prints that are just a hair over 5x7.5". To my eye this works best in monochrome, as in color the dither pattern that the printer needs to use is itself unpleasant on close inspection.

I want to emphasize that this isn't a claim about how everyone ought to make prints; it's just one way to do it, where small is preferred to large and a particular kind of subtlety is the esthetic effect that one is after. But under this very specific scenario, 360 vs 720 makes a real difference.


In case it isn't obvious, it only makes sense if the resolution of your file is higher than 360ppi without any upsampling.

YES!

I should add that using film scans as the source complicates the reckoning substantially. With traditional silver-grain monochrome film, the reason to scan at seemingly ultra-high resolutions is that it takes somewhere up toward 6000 ppi or beyond to adequately resolve film grain. Below that point you get a mealy, aliased pseudo-grain that can look really obnoxious. This is independent of the on-film resolution of the original capture.

If the enlargement ratio is small enough the problems with the grain might be suppressed below the threshold of perceptibility. But if the intent is to make big prints from, say, a scan of a 4x5 negative, without having the result look processed and poster-ish, the technical challenges are very substantial, and drum scanners show their value.

robertrose
30-Mar-2015, 17:21
What many large format photographers forget (or never understood) is that diffraction is a major limitation at small f/stops (such as f/64). Each element of the system, however, must be considered, as the loss in resolution due to diffraction may actually compensate (for example) with a deeper focus area, which could help alleviate a slight misplacement of the film plane relative to the ground glass. In other words, you need to find out what gives you the best result for your particular camera, lens, film, film holders, scanner, etc. Also, as someone else pointed out, downsizing algorithms differ, so YRMD.

All that said, we still can't neglect lighting, composition, and subject matter.

Lenny Eiger
30-Mar-2015, 19:21
What many large format photographers forget (or never understood) is that diffraction is a major limitation at small f/stops (such as f/64).

Diffraction doesn't come into play in my system until one hits f64, and a little sharpening can cover it.

There are too many variables here, the tuning of the drum scanner and everything else. Best is to do a test and see if you like things at the size one is considering. i concern myself very little with resolution these days I don't need super critical sharpness. I am much more interested int he tonality that one can get out of the film...

Lenny

paulr
1-Apr-2015, 07:22
The point at which diffraction becomes a limiting factor depends on all the other factors. The quality of the taking lens, most obviously, but I've found that the real life battle is usually between diffraction and depth of field. In other words, what's worse ... a little more defocus, or a little more diffraction? In general, if I'm striving for everything in focus, I get better results stopping down and accepting the diffraction. Its effects are even across the frame, and diffraction blur is especially easy to sharpen. Standard unsharp masking does a reasonable job on diffraction; deconvolution algorithms (like the lens blur setting in PS's Smart Sharpen filter) do an even better job.

How much quality you can recover through sharpening depends entirely on the level of noise. When the MTF of your fine detail falls below the noise floor, it's gone.

At just f22, assuming perfect optics, you're down to about 5% MTF at 75lp/mm. This is the highest resolution at which it might be possible to recover anything at all. Whether or not you can get detail that looks good, without amplifying the crap out grain, is another question. At 80 lp/mm, MTF is zero. Your decent quality detail will end at around 20%MTF, which you'll see at 55lp/mm. This corresponds to a range of 2800 to 3800 lpi scanning frequency. Anything beyond this is going to be more about resolving the grain, or oversampling to reduce noise introduced in the scanning process.

This is all about theoretical lens performance. Actual lenses are worse. Actual lenses used by actual people in the real world are considerably worse.

Here's an MTF / aperture chart for a Micro-Nikkor lens, which probably outperforms any large format optics. You'll notice the numbers are worse than what I posted above. For f32 they start to look grim:

131672

Lenny Eiger
1-Apr-2015, 10:13
Paul,

I don't know who made that chart, and I have no reason to diss anyone, but my results in my setup don't match the chart....

I have done numerous real-world tests where I have tested each one of my lenses at different f-stops, scanned them and printed them out (that last part is important). Most would say I have pretty good eyes, and I can tell you that I get excellent results from my lenses out to f45. I have Rodenstock Sironar S lenses for my 150 and 300 normal lens for the 4x5 and 8x10. Nikon lenses work just as well. I looked up the MTF for the Rodenstock's but they don't publish anything but at "optimum aperture" which they didn't specify. I also have a 150 Gold Dot Dagor which is just as good (in b&w at least).

Your numbers may be all correct, but I am not seeing this in real life examples, with my setup. Over and over again, its just not there.

I think its unfortunate that this information (true or not) keeps people shooting at open apertures and losing good depth of field. Doing a little Scheimphlug helps, but it is often misunderstood, and it doesn't help nearly as much it seems to. I scan a lot of film that comes in unsharp, from people who know their swings and tilts.

It's not grim over here...

Lenny

paulr
1-Apr-2015, 12:18
It's only grim in comparison to what's possible at lower apertures. f45 and above might still produce outstanding prints from a big neg.

Those numbers are accurate, and they represent an idealized lens. Real-world results will never actually be as goodówith the exception that sharpening can increase the MTF significantly if the S/N ratio is high enough. Keep in mind that to find the sharpness of the negative, you have to multiply the lens MTF by the film's MTF. It's going to be lower than what the lens can do. Even a digital sensor has a sloping MTF curve.

a 45lp/mm extinction resolution isn't necessarily crippling. You can make prints with contact print-like sharpness and clarity with 7 lp/mm or less at the print. I can demonstrate this all day, even with merely decent scans. This means that if you've got 30 lp/mm at 20% MTF, you'll can get outstanding sharpness even from a 4X linear enlargement. At least if you're printing digitally.

appletree
1-Apr-2015, 13:39
I do not mean to break the flow of in-depth conversation, but is this mainly only a concern for those printing from their digital files?

I never have nor intend to print from my digital files. I always scanned my 6x6 at 2400dpi and my 35mm at 3600dpi (I think...I might have that flipped). Whatever it is, I have always kept it the exact same. It only ends up on flickr or my site.

No idea yet what I was planning on scanning 4x5 in at...plus I am getting a new scanner (Epson v700). So I will have to get new film holders and everything all over again.

fishbulb
1-Apr-2015, 17:10
Yeah, I think that's the whole point. For displaying on the web, any old scanner is probably fine.

My current workflow for 4x5 is negative -> drum scan -> digital editing -> digital print. Being raised in the computer era, I find digital editing so much simpler, and darkroom work tedious at best. But some people love working in the darkroom.

Lenny Eiger
2-Apr-2015, 11:11
a 45lp/mm extinction resolution isn't necessarily crippling. You can make prints with contact print-like sharpness and clarity with 7 lp/mm or less at the print. I can demonstrate this all day, even with merely decent scans. This means that if you've got 30 lp/mm at 20% MTF, you'll can get outstanding sharpness even from a 4X linear enlargement. At least if you're printing digitally.

I have a new rule for myself. There are things that exceed the capacity of my printer and those that don't. I don't know what the lpm of the diffusion pattern is, however, I can say without a doubt that it's effect is larger than the effect diffraction, at least to f45. If I can't tell the difference in the print, then for practical intents and purposes, it is not something I need to be worried about.

I've got a very sharp scanner so where I might be losing in one area I am getting back in another. I can't imagine what an Epson scanner would do - the scanner's loss of sharpness is so dramatic to be quite a bit larger than any diffraction effect.

paulr
2-Apr-2015, 12:35
I can't imagine what an Epson scanner would do - the scanner's loss of sharpness is so dramatic to be quite a bit larger than any diffraction effect.

Larger than most diffraction effects. If you're careful, you can get at least 1800 dpi of useful resolution from an Epson scanner. This corresponds to 35 lp/mm, which, it turns out, is about the maximum useful resolution you can at f32. So at f45 and above, you notice a difference.

Lenny Eiger
2-Apr-2015, 14:23
Larger than most diffraction effects. If you're careful, you can get at least 1800 dpi of useful resolution from an Epson scanner. This corresponds to 35 lp/mm, which, it turns out, is about the maximum useful resolution you can at f32. So at f45 and above, you notice a difference.

Yeah, but notice how? Is it a 1% difference, or a 70% difference? Is it simply a loss of sharpness, does it appear like movement, or just general blurring? Can it be alleviated by a little sharpening?

I'm not using an Epson, I'm using a drum scanner with an optical resolution of 7394. The scanner won't be the limit. The lenses are excellent. The printer's dithering patten clearly will be the largest limiting factor. I'm using b&w inks, which apparently have a higher resolution, according to Cone. I'm using a high end RIP which can do all sorts of overlaying of ink for the crossover, etc.

I'm saying that I don't see any effect whatsoever until f64, not in the print. Everyone I talk to that prints, runs tests, says, yes, diffraction is a factor, but a factor small enough to be ignored.

My images are not out of focus...

Lenny