Scanning resolution, printing resolution, and downsampling

I just read Margulis' chapter on resolution and am somewhat confused
(it's one of those things you have to read multiple times, I think).
Up to now I have been scanning my 6x6 and some 35mm slides and negs at
3000-4000dpi on a Nikon LS8000 at 14-bit depth, and manipulating
mostly at full res, then downsampling to 300dpi before sending it to
my Epson 2200. According to Margulis, having a resolution too high
can soften an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much data to
the printer, and if so, wouldn't downsampling before printing correct
it, or should it be scanned at a lower resolution to begin with?

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print size,
and optimum print resolution?

Thanks for any input.

Chris

"hassy_user" wrote in message
om...
I just read Margulis' chapter on resolution and am somewhat
confused (it's one of those things you have to read multiple times,
I think). Up to now I have been scanning my 6x6 and some
35mm slides and negs at 3000-4000dpi on a Nikon LS8000 at
14-bit depth, and manipulating mostly at full res,

So far, you've done well.

then downsampling to 300dpi before sending it to my Epson
2200.

That is not optimal. The 2200 (like most desktop Epsons) internally
dithers based on a 720ppi image. If the ppi for the output size
requested is different, the printer driver will interpolate to 720
ppi. In your case, it will more than quadruple the number of pixels
you offered it. Interpolation will not help resolution.

According to Margulis, having a resolution too high can soften
an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much
data to the printer, and if so, wouldn't downsampling before
printing correct it, or should it be scanned at a lower resolution
to begin with?

As stated above, only down-sample to 720 ppi at output size. Then
sharpen the result, you can visually oversharpen a bit, because there
will be losses in the printing process. Always judge the amount of
sharpening at 100% zoom level in the photoeditor.

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print
size,
and optimum print resolution?

Scanning film at the highest native resolution will reduce apparent
graininess, so by all means keep doing it as you are. You can improve
the downsampling quality by applying a little blur before
downsampling. The amount of blur needed depends on the amount of
downsampling. That will also help reducing apparent graininess,
instead of increasing it due to grain-aliasing.
http://www.xs4all.nl/~bvdwolf/main/f...own_sample.htm
http://www.xs4all.nl/~bvdwolf/main/f...e/example1.htm
Print resolution, as stated should in your case be 720 ppi after
resampling.

Grain aliasing is explained he
http://www.photoscientia.co.uk/Grain.htm

Bart

"hassy_user" wrote in message
om...
I just read Margulis' chapter on resolution and am somewhat
confused (it's one of those things you have to read multiple times,
I think). Up to now I have been scanning my 6x6 and some
35mm slides and negs at 3000-4000dpi on a Nikon LS8000 at
14-bit depth, and manipulating mostly at full res,

So far, you've done well.

then downsampling to 300dpi before sending it to my Epson
2200.

That is not optimal. The 2200 (like most desktop Epsons) internally
dithers based on a 720ppi image. If the ppi for the output size
requested is different, the printer driver will interpolate to 720
ppi. In your case, it will more than quadruple the number of pixels
you offered it. Interpolation will not help resolution.

According to Margulis, having a resolution too high can soften
an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much
data to the printer, and if so, wouldn't downsampling before
printing correct it, or should it be scanned at a lower resolution
to begin with?

As stated above, only down-sample to 720 ppi at output size. Then
sharpen the result, you can visually oversharpen a bit, because there
will be losses in the printing process. Always judge the amount of
sharpening at 100% zoom level in the photoeditor.

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print
size,
and optimum print resolution?

Scanning film at the highest native resolution will reduce apparent
graininess, so by all means keep doing it as you are. You can improve
the downsampling quality by applying a little blur before
downsampling. The amount of blur needed depends on the amount of
downsampling. That will also help reducing apparent graininess,
instead of increasing it due to grain-aliasing.
http://www.xs4all.nl/~bvdwolf/main/f...own_sample.htm
http://www.xs4all.nl/~bvdwolf/main/f...e/example1.htm
Print resolution, as stated should in your case be 720 ppi after
resampling.

Grain aliasing is explained he
http://www.photoscientia.co.uk/Grain.htm

Bart

From: (hassy_user)

I just read Margulis' chapter on resolution and am somewhat confused

Up to now I have been scanning my 6x6 and some 35mm slides and negs at
3000-4000dpi on a Nikon LS8000 at 14-bit depth, and manipulating
mostly at full res, then downsampling to 300dpi before sending it to
my Epson 2200. According to Margulis, having a resolution too high
can soften an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much data to
the printer, and if so, wouldn't downsampling before printing correct
it, or should it be scanned at a lower resolution to begin with?

Dan is a member of the "scan for the target output" club, which is the way most
book and magazine publishers and many advertising agencies work. They rarely
print large prints (especially magazines and books) and don't spend a lot of
time fine-tuning each image, as a general rule.

The other side of the coin is the "scan once, output many" club which feels
it's best to scan at high rez, do all your prep work on a master file (which is
what you're doing) and downsample and sharpen to various output sizes ranging
from large fine art prints to tiny web thumbnails. Here's a good explanation
of the theory from a fine art lab I've used, West Coast Imaging ...
http://www.westcoastimaging.com/wci/.../scanonce.html The key
to doing this well is to know how to downsample well and how to use USM in a
variety of situations, typically with film using edge sharpening or something
similar.

You should be able to run a test yourself with the 8000 and see if downsampling
works for you ... just scan at the native rez (4,000 dpi) and downsample to say
360 ppi for a decent sized print and print it. Also scan at the exact rez for
this size print and print directly from that one too without downsampling ...
if you're printing 10x10" @ 360 ppi for example you need 3600 pixels/side so
would scan at roughly 1650 dpi (the Nikon software lets you set this exactly).

See what looks best, the 10x10" print at 1650 or the 4000 dpi scan downsampled.
Keep in mind Dan has access to much better scanners than your 8000 (which is
what I use too and he may get a better result with his 1650 dpi scan than
you or I with the Nikon. The Nikon will still scan internally at 4,000 dpi and
then downsample with its software ... can you do this better in Photoshop
yourself?

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print size,
and optimum print resolution?

You can work backwards once you know the optimal printer ppi and the desired
print size (multiply the ppi x the print dimensions to get the pixel count and
divide this by the size of the film to get the scan rez), but of course the
main problem is that you'll get a different target scan rez for each print size
and for each printer ppi. No problem if you're scanning once for a magazine
article but a hassle if you think you'll print the same file yourself at
various sizes or need even smaller images for the web.

If you want to see what a real digital ace can do with the "scan once and
resize" flow look at Bill Atkinson's work. He shoots Velvia with a Hassy and
scans at 5,000 dpi with a Tango drum, then resizes these files for anything
from a 24x24" fine art print at 360 ppi to thumbnails for the web that are
100x100 pixels ... http://www.billatkinson.com/CatalogIndex.html

Bill

From: (hassy_user)

I just read Margulis' chapter on resolution and am somewhat confused

Up to now I have been scanning my 6x6 and some 35mm slides and negs at
3000-4000dpi on a Nikon LS8000 at 14-bit depth, and manipulating
mostly at full res, then downsampling to 300dpi before sending it to
my Epson 2200. According to Margulis, having a resolution too high
can soften an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much data to
the printer, and if so, wouldn't downsampling before printing correct
it, or should it be scanned at a lower resolution to begin with?

Dan is a member of the "scan for the target output" club, which is the way most
book and magazine publishers and many advertising agencies work. They rarely
print large prints (especially magazines and books) and don't spend a lot of
time fine-tuning each image, as a general rule.

The other side of the coin is the "scan once, output many" club which feels
it's best to scan at high rez, do all your prep work on a master file (which is
what you're doing) and downsample and sharpen to various output sizes ranging
from large fine art prints to tiny web thumbnails. Here's a good explanation
of the theory from a fine art lab I've used, West Coast Imaging ...
http://www.westcoastimaging.com/wci/.../scanonce.html The key
to doing this well is to know how to downsample well and how to use USM in a
variety of situations, typically with film using edge sharpening or something
similar.

You should be able to run a test yourself with the 8000 and see if downsampling
works for you ... just scan at the native rez (4,000 dpi) and downsample to say
360 ppi for a decent sized print and print it. Also scan at the exact rez for
this size print and print directly from that one too without downsampling ...
if you're printing 10x10" @ 360 ppi for example you need 3600 pixels/side so
would scan at roughly 1650 dpi (the Nikon software lets you set this exactly).

See what looks best, the 10x10" print at 1650 or the 4000 dpi scan downsampled.
Keep in mind Dan has access to much better scanners than your 8000 (which is
what I use too and he may get a better result with his 1650 dpi scan than
you or I with the Nikon. The Nikon will still scan internally at 4,000 dpi and
then downsample with its software ... can you do this better in Photoshop
yourself?

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print size,
and optimum print resolution?

You can work backwards once you know the optimal printer ppi and the desired
print size (multiply the ppi x the print dimensions to get the pixel count and
divide this by the size of the film to get the scan rez), but of course the
main problem is that you'll get a different target scan rez for each print size
and for each printer ppi. No problem if you're scanning once for a magazine
article but a hassle if you think you'll print the same file yourself at
various sizes or need even smaller images for the web.

If you want to see what a real digital ace can do with the "scan once and
resize" flow look at Bill Atkinson's work. He shoots Velvia with a Hassy and
scans at 5,000 dpi with a Tango drum, then resizes these files for anything
from a 24x24" fine art print at 360 ppi to thumbnails for the web that are
100x100 pixels ... http://www.billatkinson.com/CatalogIndex.html

Bill

Bill Hilton wrote:


Hi...

much snipped

You should be able to run a test yourself with the 8000 and see if downsampling
works for you ... just scan at the native rez (4,000 dpi) and downsample to say
360 ppi for a decent sized print and print it. Also scan at the exact rez for
this size print and print directly from that one too without downsampling ...
if you're printing 10x10" @ 360 ppi for example you need 3600 pixels/side so
would scan at roughly 1650 dpi (the Nikon software lets you set this exactly).

more snipped

For what it's worth, I did this test using many
variations... about 20 prints worth. (Kodak machine
in the mall type prints)

Blind tests for which was better was always extremely
close. Testers (included print machine operator, the
subject herself, neighbors and friends) after long
consideration always concluded that the "big" scan,
untouched by PS or PSP, was the best.

Ken

Bill Hilton wrote:


Hi...

much snipped

You should be able to run a test yourself with the 8000 and see if downsampling
works for you ... just scan at the native rez (4,000 dpi) and downsample to say
360 ppi for a decent sized print and print it. Also scan at the exact rez for
this size print and print directly from that one too without downsampling ...
if you're printing 10x10" @ 360 ppi for example you need 3600 pixels/side so
would scan at roughly 1650 dpi (the Nikon software lets you set this exactly).

more snipped

For what it's worth, I did this test using many
variations... about 20 prints worth. (Kodak machine
in the mall type prints)

Blind tests for which was better was always extremely
close. Testers (included print machine operator, the
subject herself, neighbors and friends) after long
consideration always concluded that the "big" scan,
untouched by PS or PSP, was the best.

Ken

(hassy_user) wrote in message . com...
I just read Margulis' chapter on resolution and am somewhat confused
(it's one of those things you have to read multiple times, I think).
Up to now I have been scanning my 6x6 and some 35mm slides and negs at
3000-4000dpi on a Nikon LS8000 at 14-bit depth, and manipulating
mostly at full res, then downsampling to 300dpi before sending it to
my Epson 2200. According to Margulis, having a resolution too high
can soften an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much data to
the printer, and if so, wouldn't downsampling before printing correct
it, or should it be scanned at a lower resolution to begin with?

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print size,
and optimum print resolution?

Thanks for any input.

Chris

Thanks Bart, Bill, and Ken for the thoughtful responses. Bill and
Bart, you apparently differ on the recommended output ppi (360 vs
720). Will there be a visible difference, however subtle, in choosing
one over the other?

Chris

(hassy_user) wrote in message . com...
I just read Margulis' chapter on resolution and am somewhat confused
(it's one of those things you have to read multiple times, I think).
Up to now I have been scanning my 6x6 and some 35mm slides and negs at
3000-4000dpi on a Nikon LS8000 at 14-bit depth, and manipulating
mostly at full res, then downsampling to 300dpi before sending it to
my Epson 2200. According to Margulis, having a resolution too high
can soften an image, so I'm trying to find out two things:

1. Is the softening that occurs an effect of sending too much data to
the printer, and if so, wouldn't downsampling before printing correct
it, or should it be scanned at a lower resolution to begin with?

2. Where can I find a reference of starting points for optimum
scanning resolution that considers input film size, final print size,
and optimum print resolution?

Thanks for any input.

Chris

Thanks Bart, Bill, and Ken for the thoughtful responses. Bill and
Bart, you apparently differ on the recommended output ppi (360 vs
720). Will there be a visible difference, however subtle, in choosing
one over the other?

Chris

"hassy_user" wrote in message
om...
SNIP
Bill and Bart, you apparently differ on the recommended output
ppi (360 vs 720). Will there be a visible difference, however
subtle, in choosing one over the other?

Depends on the image processing done to the final sized file, and
whether it had enough resolution to begin with. Assuming the file has
enough native resolution, the 720 ppi version will allow to enhance
edge contrast more accurately. The resulting output will look better
in that case.

If you use a dedicated print program like Qimage, life becomes even
easier because it takes care of it automagically (including profile
conversion and optimization of paper used), without altering the
original file.

Bart