Soft scans or soft images?

[walter23]

Attached is a pretty typical transparency scan (in terms of sharpness). Scanned at 1200 DPI on my epson V750. Second is a full 1200 DPI crop. Is this normal for this kind of scan? I'd like to get slightly sharper results, but I can't decide if my in-camera technique (or lenses) are failing or my scanner is at fault. It just seems to me that these shouldn't really be falling apart at only 1200 DPI.

I can think of a couple of very obvious variables in camera that could lead to soft images. 1. I use a fairly stiff cable release and my exposures often end up being 2 second "T" exposures. 2. I stop down to f/45 or f/64 for a lot of my shots to ensure adequate depth of field but could diffraction be the cause? 3. Lens sharpness? (caltar II-N rodenstock lenses - 90 and 210 at the moment).

The conundrum is that I have to examine my transparencies with my scanner. I'm happy with the sharpness of my 11x14 and 16x20 B&W prints from the darkroom; I just haven't printed any colour yet (nothing I'm happy enough with to print).


(No criticism on the aesthetic of the image please. It's not a good shot, I know. Just happened to be convenient for this)

[Rob Champagne]

an inch on your screen will be something like 96dpi. An inch in a print is something like 360dpi or maybe 720dpi. That means looking at it on screen is roughly equivalent to a 4 times (or 8 times depending on your chosen print resolutoin) enlargement to how it will look when when printed. Given that scans will be a little soft anyway and your aperture sounds to be a little small for optimum sharpness and your lens may not be the sharpest out there, then it is hardly surprising that what you see on screen is not sharp. It never will be. Just how unsharp is depnds on above factors.
You best option is to print a small test patch and judge that and not your screen image.

[Rob Champagne]

and when you scan at at 1200dpi when your flatbed scanner is capable of much higher resolution, you leave big gaps between each dot which ruins the sharpness. The dot size is not variable on a flatbed in the same way it is on a drum scanner.
Scan at native scanner resolution and then downsize by factors of 2 to get to where you want to be.

e.g. say you want to make a 16x20 print printed at 360dpi. Your file size needs to be 5760x7200pixels.

at the scanners 4800dpi setting, that gives you a file size of 19200x24000. So first downsize would be to 11520x14400 (twice final size) and the second downsize would be to 5760x7200.
Apply a little sharpening before each dowsize. I use 50% 0.7 others mileage may differ and some do the downsize in one go. The downsize loses many small scan artefacts and tiny dust specs. Best to spot it before the first dowsize and don't don't sharpen after last downsize unless you really need to.

Now I can hear you say, but I don't have the RAM for this and it takes a long time for the scan. Well if thats the case, then why didn't you buy a scanner with a larger scan dot size which equates to a lower resolution scanner, because by buying that top of the line epson and not scanning at its native resolution, you are throwing away its capabilities.

[walter23]

Are you saying the scanner doesn't sort of downsample on the fly but rather just discards every nth pixel in its linear sensor?

[Rob Champagne]

Yup! If it were downsampling it would be even slower. But usually when you scan at lower res its a hell of a lot faster which wouldn't be the case if it were scanning at native resolution and then doing the additional process of downsampling.

[walter23]

If the scanner firmware did the downsampling before sending the data to the computer you'd avoid the communication bottleneck which might account for the speed increase... but I'll give your suggestion a try. It rings true because I've noticed that my higher resolution scans look better (and not just because of the obvious higher resolution). I figured maybe it was switching to a different sensor or optics (based on something someone once said on a forum I think), or maybe using finer control over the stepping motor. But when I think about it those explanations don't make much sense.

[Rob Champagne]

For absolute best print quality scan at 6800dpi and adjust print dpi only to obtain print size you want. i.e. no downsize. The higher print res will compress more dots into an inch and retain more original fine detail. I know some people will say you can't see the difference but I urge you to try it if you have the memory available to do it.

[Greg Lockrey]

This has be very enlightening, Rob. I'll have to look at my film scanning with the 750V differently from now on.

[Aender Brepsom]

Rob, shouldn't you be nicknamed "An old dog teaching new tricks"? :D
I'd like to thank you for the precious information. I guess they are valid for other flatbed scanners too. I'll have to try it.
Really helpful! Cheers

[Ted Harris]

I have to disagree with scanning at 6800 spi. It is pure overkill and what you are actually doing is sapling a lot of data that contains no information. The problem being that the true optical resolution of this scanner is more like 2200 spit according to our tests and the tests of others. So, if you sample at 2400 spi you are capturing all the information on the film. Not to mention that a 6800 spi scan of a full frame 4x5 tranny will give you a really huge file that you will find difficult to impossible to manipulate unless you have a powerful computer system. In fact, when I tested this scanner when it first came out I did try some scans at 6800. Tried some specifically on a MacBook with 2 GB of RAM to see what would happen ... the system consistently crashed. When I scanned on the G5 Tower with 5 GB of RAM all was well if awfully slow.

[JPlomley]

The conundrum is that I have to examine my transparencies with my scanner.

I would invest in the Schneider 3x loupe designed for 6x7cm. It will allow you to critically evaluate your chromes and weed out soft images.

Try to avoid f/45 and f/64. In my experience I try not to shoot past f/32.3 (3mm focus spread). It's also important to characterize your lenses to see what you can get away with. A simple high frequency subject imaged at f/22, f/32, f/45, and f/64 will quickly reveal a deterioration in image quality. For example, I know my 135, 210, and 300 APO Sironar-S lenses can handle f/45 if absolutely necessary, and still provide excellent image quality. However, f/45 is unacceptable on the 75 and 90 Grandagon-N's, so I will optimize my movements/composition to avoid the focus spreads requiring this aperture on the wide angles.

Have to agree with Ted on the sampling resolution having gone through this myself. All you are doing is accentuating film grain, thereby reducing your signal to noise ratio. The advantage of shooting 4x5 is that we can scan at lower resolution compared to the smaller formats for equivalent print size, thereby eliminating noise. If you are making exceptionally large images and therefore need to scan at maximum resolution for the pixel count, then you better plan on running a noise reduction program like Neat Image or Noise Ninja. But on a file that size, you can take a long break for lunch.

[Greg Lockrey]

Based on what Rob and Ted are both saying, then the optimal resolution for film scanning with the 750V would be at the 2400 ppi setting since the scanner's max is really 2400 vs stated 4800 ppi. (How does one determine this?) Down sizing the file from there would work too I would imagine. Another point about grain size and scanning was made about too high of a ppi would inhance the grain. I always thought that if the pixal size was larger than the grain, then the grain seemed to have increased to the pixal dimension. That was why scanning at higher resolutions was better to minimize grain. I'm confused now.:confused:

[Ted Harris]

Greg, I test scanners using an AIG test target, these are the imaging industry standards.

[Greg Lockrey]

Greg, I test scanners using an AIG test target, these are the imaging industry standards.

I was just curious, thanks. We don't always get access to this high tech stuff here in the "Tundra". ;) Are there links to results? Or does the 96 ppi of a monitor screw up the results too much?

[Brian Ellis]

Agree with Ted. Scanning at a higher ppi than the scanner is capable of resolving does nothing but increase the file size without any gain in information.

As to the original question, there are so many possible variables here it's hard to tell exactly why your cropped image is soft. Clearly you should be scanning at about 2200 ppi rather than 1200. But apart from that, it sounds like you might just be more or less automatically stopping down to f45 and f64 to gain depth of field. If that's the case you might read the article in the articles section of this forum dealing with focusing the view camera. With a little study of that article you'll learn how to find the optimum aperture for any given scene, which often will be less than f45 or f64 unless you're in the habit of making photographs with extreme near-far relationships. Among other benefits, that will allow you to increase shutter speeds to something faster than the several seconds you mention. Using a faster shutter speed obviously will reduce the effects of slight camera movements, subject movement, etc.

[Gene McCluney]

Something that hasn't been mentioned specifically in this thread, but it seems that ALL scanners require a bit of "sharpening" of the scan in Photoshop, after you size the image to your final print size.

[Ted Harris]

I think we all may have been assuming that the OP knew about sharpening, but maybe not. In fact , for consumer scanners it is generally more than a bit of sharpening that is required. As a rule of thumb you can start with a radius that is 5% of your resolution and 220% and then back off until your image looks the way you what it too; that is, sharp enough but without obvious digital artifacts. Also remember that you must sharpen at the printing size or things may get interesting.

[walter23]

"3mm focus spread"

Can you explain this terminology? I've never heard of this and I can't even guess.

Sharpening - yeah, I know about that. Of course the thing about sharpening as we all know is that you don't gain any real resolution, just increase the perception of edges. I tend to err on the side of undersharpening to avoid haloing.

[65Galaxie]

You just need to add some sharpening in photoshop. Flatbed scanners show very blurry without any sharpening. Here's a slight redo on your crop. The extra grain won't be noticable on enlargening to at least 16x20.

[Brian Vuillemenot]

I wouldn't stop down past f/32 routinely, except if you need the extra depth of field, by all means stop down rather than get out of focus areas. Be sure to use appropriate unsharp masking. The Epson flatbeds are really only good for 3-4X enlargements, so if you are making bigger ones, you may want to invest in drum scans. It's not your lenses- I have a 210 Caltar-IIN, and it's among my sharpest lenses- definately as good as my 150 Apo Sironar-S, and probably sharper than my 110 SS-XL.

[Kirk Gittings]

The problem being that the true optical resolution of this scanner is more like 2200 spit according to our tests and the tests of others. So, if you sample at 2400 spi you are capturing all the information on the film. Ted Harris

Ted's experience mirrors my own independent tests on the 750, 2400 is really closer to the optical resolution of this scanner and there is no advantage and considerable more tediousness by scanning above this.

Also.....I have been doing some testing on this lately, though I do not print large, and this is by no means a recomendation at this point.......... It appears to me if you need a larger file it may be better in terms of actual print resolution to scan at the optical 2400 resolution, and up res with Genuine Fractals (the latest version has some real world improvements) later in the workflow before the final sharpening.

[QT Luong]

Can you explain this terminology? I've never heard of this and I can't even guess.



http://www.largeformatphotography.info/fstop.html

[walter23]

You just need to add some sharpening in photoshop. Flatbed scanners show very blurry without any sharpening. Here's a slight redo on your crop. The extra grain won't be noticable on enlargening to at least 16x20.

That's definitely something of an improvement, however like I said I tend to prefer to undersharpen rather than oversharpen and I want to figure out how much of the softness is from my camera technique and how much is scanner. The output you're seeing there has some auto-sharpening from the scanning software already applied.

I think what I'll do though is take this tranny to the darkroom* and check it with the grain focusing tool. That should tell me all that I need to know (and I'll report back once I get to it. Alas I'm swamped so it won't be for a week or so).



*I've got no other reason to ever have a tranny in a dark room, I assure you.

[Ted Harris]

The output you're seeing there has some auto-sharpening from the scanning software already applied.



If you are using auto sharpening that could also be part of the problem. I suspect that the errors are all in some combination of the scanner and your scanning workflow. Regardless of whether you are using Epson's software or Silverfast you should never just use the auto settings.

Time for a small bit of self promotion. Our next scanning and digital workflow workshop will be at Midwest Photo in Columbus in late January or early February. Kirk and I will be doing the workshop. There will also be one in late February or early March at Elevator in Toronto. Working on dates now. Send me a PM if you want to be on the mailing list

[Kirk Fry]

Seems to me the first stop on this fuzzy hunt would be to start at the beginning. Look at the developed film with a good optical magnifying device of some sort (Loupe, dissecting scope, microscope.) and figure out if it is the film that is fuzzy or the aftermath.

K

[Rob Champagne]

That's definitely something of an improvement, however like I said I tend to prefer to undersharpen rather than oversharpen and I want to figure out how much of the softness is from my camera technique and how much is scanner. The output you're seeing there has some auto-sharpening from the scanning software already applied.

I think what I'll do though is take this tranny to the darkroom* and check it with the grain focusing tool. That should tell me all that I need to know (and I'll report back once I get to it. Alas I'm swamped so it won't be for a week or so).

*I've got no other reason to ever have a tranny in a dark room, I assure you.

I'm not going to argue the points others have made but I would remind you that you have already said that you have observed better quality from higher resolution scans.
Ultimately you should do your own print tests using the different techniques given here and trust your own vision of what you actually see rather than a theory or non scientific test.

[Greg Lockrey]

I'm not going to argue the points others have made but I would remind you that you have already said that you have observed better quality from higher resolution scans.
Ultimately you should do your own print tests using the different techniques given here and trust your own vision of what you actually see rather than a theory or non scientific test.

I think that what they are saying is that the limit for resolution is met about half way with the 750V in terms of ppi. But I wonder if "throwing out" half the pixal potential because the lens can't resolve them anyway is a good thing too? I'll have to play with this a bit to see for myself. (What if I have that lens that is on the high end of the production statistical curve?)

I know from experiance and testing a step wedge of the 256 gray points what the limits are for my printers and paper combinations. Trying to go beyond laws of physics in this case is just a waste of ink. Getting the max resolution will save time by not needing to go over.

[walter23]

Just as a followup, I checked this out in the darkroom and under the grain focuser the rocks are sharper. I'm using a decent enlarger lens (schneider componon-S) and I can barely make out the transparency grain. A couple of other transparencies that I've scanned, e.g. this one, are really sharp but don't show quite as nice detail in the scans:

http://ashphotography.ca/zenphoto/albums/landscapes/boom-lake-trail-creek.jpg

I guess I'm just hitting the limit from this kind of scanner. Might be able to adjust the height of the transparencies a bit to improve things.

It is, however, encouraging that I could see almost the same detail in the scans as I can see with the grain magnifier under my enlarger, even if the scans are a bit softer.