Help: Scanning films at highest DPI but original size?

[Best]

Hi, After reading a lot up on scanning, I'm getting closer to finally
starting to scan all my film negatives. But I want to be right, or as
correct as possible, from the start.

With my Epson 3170, It can scan 35mm negatives at 1200dpi, the highest
resolution it's capable. On top of that, I can choose to scan at the
negative's original size, roughly 1.5 in by 1 in, or a page size of 8.5
in by 11 in. In the second case, would the scanning software be just
adding artificial pixels to the scan at the original size? Since I'm
doing archiving, adding artificial pixels isn't necessary and it wastes
space. I can do that later in Photoshop if I need to.

So again. With the resolution at 1200dpi, is scanning at a size larger
than the original size produce a better image?

Excuse me for my not-so-technically-correct language.

 

[Wayne Fulton]

Hi, After reading a lot up on scanning, I'm getting closer to finally
starting to scan all my film negatives. But I want to be right, or as
correct as possible, from the start.

With my Epson 3170, It can scan 35mm negatives at 1200dpi, the highest
resolution it's capable. On top of that, I can choose to scan at the
negative's original size, roughly 1.5 in by 1 in, or a page size of 8.5
in by 11 in. In the second case, would the scanning software be just
adding artificial pixels to the scan at the original size? Since I'm
doing archiving, adding artificial pixels isn't necessary and it wastes
space. I can do that later in Photoshop if I need to.

So again. With the resolution at 1200dpi, is scanning at a size larger
than the original size produce a better image?

I dont have that scanner, so I cant see what you are seeing, but the
Epson 3170 scanner is rated as a 3200 dpi scanner, not 1200 dpi. It
should work at 3200 dpi, at least for small things like 35 mm film.
The only purpose of 3200 dpi is to enlarge small film, like 35 mm size
film.

But, the short answer to the question asked is that the scanning
resolution you enter in the dialog box is NOT the scanning resolution.
It is instead the printing resolution (this is how all scanners work).
The actual scanning resolution (sampling resolution) is the product of
Resolution x Scale that you enter.

So, you did not say what scale factor you are using, but you said page
size of 8.5x11 inches from 35 mm film, which sounds like it must be
about 900%. If you are entering a large scale factor, like 900% size,
AND also entering a large value like 1200 dpi, then this is surely the
problem you are seeing. It is being told to scan at 1200x9 = 10800 dpi,
and that is a really serious problem for a 3200 dpi scanner.

Because, here are examples of how things actually work:

If you enter 300 dpi, 100% scale, 300x1 is 300, so it scans at 300 dpi,
and the image is scaled to print original size at 300 dpi, because this
is what you asked for.

If you enter 300 dpi, 400% scale, 300x4 is 1200 dpi, so it scans at
1200 dpi, and the output image is scaled to print 4x size at 300 dpi.
For 35 mm film, 4x full frame is a little less than 6x4 inches.

If you enter 1200 dpi 200% scale, 1200x2 is 2400 dpi, so it scans at
2400 dpi, and the image is scaled to print 2x size at 1200 dpi.
Printing photos at 1200 dpi is not a good thing to do, but in this
case, it is what we asked it to do. Probably what you meant instead
was 300 dpi 200% scale, or perhaps even 300 dpi 800% scale (scans at
2400 dpi).

If scanning 35 mm film, if you enter 300 dpi (printing resolution), and
if you enter say 900% scale (prints 9x size of 35 mm film, which is
about 8x12 inches), then it scans at 300x9 = 2700 dpi, and outputs the
image scaled to print 9x size at 300 dpi (because this is what we asked
it to do. This is the natural thing to do).

But odd (non-standard) resampled values like 2700 dpi (not integer
divisors of the 3200 dpi optical rating) are really not the best
results, so you could instead scan small 35 mm film at say
1600 dpi 100% scale and it will scan at 1600x1 = 1600 dpi, or
2400 dpi 100% scale and it will scan at 2400x1 = 2400 dpi, or
3200 dpi 100% scale and it will scan at 3200x1 = 3200 dpi,
and then you can scale it yourself later to print how you wish (300
dpi, or however that number of pixels will print).

Or you could enter 250 to 300 dpi (printing resolution) and the scale
factor of how large you want the print to be, like 900%.

These are the two ways it can work. 1200 dpi 900% wont work.

For for any scale factor except 100%, the dpi entered is the printing
resolution, NOT the scanning resolution. True at 100% too, but there
is the natural equivalency at 100% or 1x.

The site below is about basics.

 

[Best]

But odd (non-standard) resampled values like 2700 dpi (not integer
divisors of the 3200 dpi optical rating) are really not the best
results, so you could instead scan small 35 mm film at say
1600 dpi 100% scale and it will scan at 1600x1 = 1600 dpi, or
2400 dpi 100% scale and it will scan at 2400x1 = 2400 dpi, or
3200 dpi 100% scale and it will scan at 3200x1 = 3200 dpi,
and then you can scale it yourself later to print how you wish (300
dpi, or however that number of pixels will print).

Thanks for another lession well taught.

I intend to do as you said here. But 1600 is not listed on the
scanner's drop-down list. It goes 600, 720, 800, 1200, then leaps to
3200. However I can munually enter it. Do you think 1600 dpi might be
some odd number for this Epson 3170 because it's not listed?

 

[Wayne Fulton]

Thanks for another lession well taught.

I intend to do as you said here. But 1600 is not listed on the
scanner's drop-down list. It goes 600, 720, 800, 1200, then leaps to
3200. However I can munually enter it. Do you think 1600 dpi might be
some odd number for this Epson 3170 because it's not listed?

The Epson web site specs say the 3170 has "Optical Resolution" of 3200
dpi. Meaning, that is how the sensors on this model are built.
Generally, integer divisors of that value give the best quality (divided
by 1, 2, 3, 4, 8, etc). That would certainly include 1600 dpi, being
3200/2. So 1600 dpi would seem to be an excellent value, if it serves
the purpose.

The reason this is best is because the scanner has sensors that size, so
it might use every one, or every other, or every third, for an exact
result. But to resample the image to give any other value, like say 1500
dpi size would be very awkward, and it is best to skip that resampling
step. It depends really, some scanners resample better than others, but
still, no question that the photo editor can do it best. It is not at
all fatal or end of the world, but still best not to when possible, at
least some think so. Others dont notice.

Resolution really depends on your goal, your purpose for the scanned
image, how large do you need it to be to do what you want to do?

For example, 800 dpi 100% should be enough image size to show 35 mm film
on the video screen. However 1600 dpi 100% is not enough pixels to
enlarge 35 mm film to print 8x10 size, if that is the goal. The ratio
of (scanned res/printing res) is the printed enlargment factor. 8x12
inches is roughly 9x from 35 mm, so we seek a ratio that gives 9x. For
example 2700/300 is 9x and does the size job, so 2700 dpi is the
ballpark, but is not an integer divison for this scanner. It will work,
but we have more choices too. The 300 dpi is not a magic value either.

Still speaking of 35 mm film size, 1600 dpi could print 8x12 inches at
1600 dpi / 9x = 178 dpi (approx, full frame, no cropping). 178 dpi is
not so terrible really, but we would prefer it to be higher, if possible
250 dpi or more. 3200 dpi would print (assuming 8x12 from full frame 35
mm image) at about 3200/9x = 355 dpi. That would be about my upper
limit.

Or 3200 dpi gives 10x at 320 dpi, or 11x at 291 dpi. Meaning, if you
crop the image a little tighter for a better picture, to be a little
smaller area, then the 9x increases slighty to still have 8x12 inches,
so that reduces this 355 dpi number somewhat (it assumed full frame).

The way I would do it (assuming printing 8x10 inches from 35 mm film is
the realistic goal) is to routinely scan at 3200 dpi 100%.

Then in the photo editor, I would finish it up, and crop it as desired
(cropping sometimes works magic to enhance images, getting rid of dead
space, emphasizing the proper subject more - you can experiment with
different "views" in the photo editor), then scale that final result to
the 8x10 size for printing. Depending on cropped size, this 3200 dpi
case might come out 355 dpi, 307 dpi, 283 dpi, it doesnt really much
matter, but that ballpark is good. I think this is the best way.

 

[Best]

The Epson web site specs say the 3170 has "Optical Resolution" of 3200
dpi. Meaning, that is how the sensors on this model are built.
Generally, integer divisors of that value give the best quality (divided
by 1, 2, 3, 4, 8, etc). That would certainly include 1600 dpi, being
3200/2. So 1600 dpi would seem to be an excellent value, if it serves
the purpose.

The reason this is best is because the scanner has sensors that size, so
it might use every one, or every other, or every third, for an exact
result. But to resample the image to give any other value, like say 1500
dpi size would be very awkward, and it is best to skip that resampling
step. It depends really, some scanners resample better than others, but
still, no question that the photo editor can do it best. It is not at
all fatal or end of the world, but still best not to when possible, at
least some think so. Others dont notice.

Resolution really depends on your goal, your purpose for the scanned
image, how large do you need it to be to do what you want to do?

For example, 800 dpi 100% should be enough image size to show 35 mm film
on the video screen. However 1600 dpi 100% is not enough pixels to
enlarge 35 mm film to print 8x10 size, if that is the goal. The ratio
of (scanned res/printing res) is the printed enlargment factor. 8x12
inches is roughly 9x from 35 mm, so we seek a ratio that gives 9x. For
example 2700/300 is 9x and does the size job, so 2700 dpi is the
ballpark, but is not an integer divison for this scanner. It will work,
but we have more choices too. The 300 dpi is not a magic value either.

Still speaking of 35 mm film size, 1600 dpi could print 8x12 inches at
1600 dpi / 9x = 178 dpi (approx, full frame, no cropping). 178 dpi is
not so terrible really, but we would prefer it to be higher, if possible
250 dpi or more. 3200 dpi would print (assuming 8x12 from full frame 35
mm image) at about 3200/9x = 355 dpi. That would be about my upper
limit.

Or 3200 dpi gives 10x at 320 dpi, or 11x at 291 dpi. Meaning, if you
crop the image a little tighter for a better picture, to be a little
smaller area, then the 9x increases slighty to still have 8x12 inches,
so that reduces this 355 dpi number somewhat (it assumed full frame).

The way I would do it (assuming printing 8x10 inches from 35 mm film is
the realistic goal) is to routinely scan at 3200 dpi 100%.

Then in the photo editor, I would finish it up, and crop it as desired
(cropping sometimes works magic to enhance images, getting rid of dead
space, emphasizing the proper subject more - you can experiment with
different "views" in the photo editor), then scale that final result to
the 8x10 size for printing. Depending on cropped size, this 3200 dpi
case might come out 355 dpi, 307 dpi, 283 dpi, it doesnt really much
matter, but that ballpark is good. I think this is the best way.

I wish I could scan all my stuff at 3200 dpi, but it takes a long time
to scan and it takes a bigger disk space. So I'm only doing that for
some real good shots. The rest are primarily family photos to be viewed
on screen only. Scan at 1600 dpi and save in tiff should be good
enouhg, right?

Another question with previewing. I'm using Epson Scan v2.32a, the
latest version. When I draw a marquee around a negative, the color tone
inside the marquee changes depending on how much white or black space
around the negative that I include. The tone can go from a darkish blue
to a bright orange. Once I click scan, the image comes out would also
have a tone ranges from darkish blue to orange, depending on how the
preview looks. However, with whatever tone in the output, I can use
Photoshop's tools to modify it to something close to what I want.
Mainly with Levels and Color balance. I just don't know if I'm having
the wrong scan to start with.

With Silverfast Ai demo, this happens too, but much minor. Sometimes
it's more of a change in brightness for the selected area.

Because of this, how should I use the marquee to select the scan area in
preview to get the best initial scan to bigin with?

 

[Wayne Fulton]

I wish I could scan all my stuff at 3200 dpi, but it takes a long time
to scan and it takes a bigger disk space. So I'm only doing that for
some real good shots. The rest are primarily family photos to be viewed
on screen only. Scan at 1600 dpi and save in tiff should be good
enouhg, right?

Image size always depends on the purpose for that image of course, that is
the only answer possible. There is no one general answer possible.

Assuming we are still speaking about scanning 35 mm film, the 3200 dpi size
would be useful for the 9x 8x12 inch enlargement of 35 mm film. If you're
not doing that, then its excessive for lesser purposes. I doubt many of us
routinely print everything 8x10. When we do this, its probably more a
special project on the rare exception.

But I'd say yes, 1600 dpi should routinely print 6x4 inch prints fine (4.2x
enlargement from full frame 35 mm film), even after moderate cropping.
This is about 6MB in a TIF file, and about a 2200x1500 pixel image,
comparing in size to 3 megapixel camera.

But if you wont be printing 6x4 inches either, and many of us dont, then
800 dpi size should be around 1100x750 pixels, full frame, surely enough
for full screen video viewing.

Many people have the idea that they should scan and archive the hugest
possible image of absolutely everything, for any "just in case" situation.
However, if time and disk space are factors, then I think it is better to
have realistic goals, suited for the 95% instead of suited for the 5%. If
scanning the past 30 years of slides which we have not even viewed in 10
years, then the need to print 8x10 prints of every one seems very remote to
me. The video screen size seems the realistic goal to me. Then make
exceptions as you go for the special ones. The few times this might fail,
then you can come back when needed.

In practice, at least when scanning our own film, we know at the time of
first scanning if this one frame is in that special 5% deserving special
interest, and we dont have to scan every one the same size, and we dont
even have to come back later, we already know about this one, as a rule.
But everyone has different notions.

Another question with previewing. I'm using Epson Scan v2.32a, the
latest version. When I draw a marquee around a negative, the color tone
inside the marquee changes depending on how much white or black space
around the negative that I include. The tone can go from a darkish blue
to a bright orange. Once I click scan, the image comes out would also
have a tone ranges from darkish blue to orange, depending on how the
preview looks. However, with whatever tone in the output, I can use
Photoshop's tools to modify it to something close to what I want.
Mainly with Levels and Color balance. I just don't know if I'm having
the wrong scan to start with.

With Silverfast Ai demo, this happens too, but much minor. Sometimes
it's more of a change in brightness for the selected area.

Because of this, how should I use the marquee to select the scan area in
preview to get the best initial scan to bigin with?

I'm not sure I understand the question, but the marquee area you mark to
scan should only include the actual photo area you want to scan, that is,
it should not include any of the blank film borders, sprocket holes, etc.
The purpose of marking the marquee is to mark the area we want to scan, the
actual photo area to include in the image. Crop at the preview so to speak,
at least to get it inside the frame size, to include only actual photo
area. You dont really want those unnecessary areas influencing the
contrast adjustments it is making. It uses the content of the marked area
to compute the scan parameters, so to speak, so you want it all to apply to
the real world.

I suppose the blue/orange comments refer to that blank film area, which in
color negatives is orange, which is cyan or blue when inverted to positive.
But you dont want any of that blank area in your scan area, that's not the
part that we want to consider.