Page 1 of 1

Scan image image pixel dimension too small?

PostPosted: Mon Aug 03, 2009 5:15 pm
by hassiman
I don't mean to as they say "beat a dead horse" but I am still a bit uncertain as to how to get the results I want from SilverFast AI with my CoolScan 9000.

As unrealistic as it might seem, I have used as my standard of quality results a B&W 6X6 T-Max negative which I had scanned for me by the master Mac Holbert from Nash Editions.

When I ordered the scan I told him that I wanted it scanned so that it would print to 30X30 or larger if necessary with high quality.

The resultant scan, which was made on a Scitech has an image size of 11,454 X 11,401 pixels and is 124.6 MB in size and it is quite incredible.

Now using Silverfast AI on my SuperCoolSacn 9000 with the scanning method which Karl-Heinz from SilverFast recommended ( Set Q Factor at 1.5 SCREEN at 152 leave INPUT & SCALE% at default then put the size print you intend to make in inches in the OUTPUT fields.. Doing this changes the SCALE% and the DPI) I get a scan from a full-frame 6X6 color image that is only 3,752 X 3,767 pixels with a file size of 91.2 MB. Karl says this method will yield good prints and avoid the 9000 interpolating the image data.

I understand that it would be completely unrealistic to expect results from my 9000 that equal an $80,000+ SciTech... but if I want to print large shouldn't my pixel dimensions be bigger than 3,752 X 3,767... ? Or does pixel dimension have nothing to do with how large the final image is to be printed?

How do I set SilverFast AI to yield larger dimension scan files in pixels without damage by interpolation? Pror to the final scan of this image I checked to see what actual optical resolution the scanner was achieving when set up this way and it indicated that the CoolScan 9000 was using only 1,333 dpi of a supposed 4,000 of DPI.

Am I missing something here? :(

Re: Scan image image pixel dimension too small?

PostPosted: Wed Aug 05, 2009 7:54 am
by LSI_Heidorn
Dear hassiman,
if you are scanning from a 6x6 cm original and the output should be 30x30 inch( not cm, right ? ) i would propose the following:
-set Q-Factor to 1.0 as you are probably printing on an Ink-Jet / dye sub system, not giving it to Offset printing, rght ?
-set the screen to 300 dpi
-set the output fields to 30 x 30 inch.

This should result in a scan of approx. 9000 x 9000 pixel = 500 MByte ( if scanned in 48 Bit Color, if using 24 Bit -> 250 MByte ).
May finetune the USM after applying these settings or use the automatic USM setting to your liking.

best regards,

Nils Heidorn

P.S. i am assuming 30x30 inch, probably Karl-Heinz did the calculation for 30x30 cm but the enourmous size of the Scitech scan looks to me as if you intended 30x30 inch...
The only other thing i am doing is to use 300 dpi as the output screen as compared to 1,5 x 152 = 228 Output screen, so i am a bit more aggressive there... Also do not forget th care for a good USM setting, this is important for printouts in large sizes !

Re: Scan image image pixel dimension too small?

PostPosted: Wed Aug 05, 2009 3:02 pm
by hassiman
Thanks Nils,

If I want to print for the largest size my Epson 3800 can make IE 16 X 20 from 35mm or 17 X 17 from 6X6 and seeing that the 3800 Epson prefers 360 dpi rather than 300 dpi what settings should I use?

Thnks,

Rich

Re: Scan image image pixel dimension too small?

PostPosted: Thu Aug 06, 2009 8:01 am
by LSI_Heidorn
Dear rich,
as you mention the 3800 i am now sure you are talkink of inches not of cm :-)

Again, leave the calculation to the Software, you just set Q Factor to 1.0 and the screen to the desired output res ( e.g. 360 or 300 ) and fill in the output size ( e.g. 17" x 17" ) and the rest will be calculated for you, its that simple !
In the early days i also scanned to 360 dpi for epson as the drivers were not so good in interpolating but now i dount that you will see the difference if you scan to 300 or to 360 dpi, it will just save you a lot of time & file size if you use 300 dpi, so thats my choice but feel free !


So-> let SilverFast do the maths :-)

( well the outcome will be: 6 cm = 2.36 inch, you go from 2,36 inch to 17 inch = 720 %. So the input resolution will be 300 dpi * 720% = 2160 dpi )

Greetings,

Nils Heidorn

Re: Scan image image pixel dimension too small?

PostPosted: Thu Aug 06, 2009 7:24 pm
by hassiman
Dear Nils...

Given your response let me go through the actual steps as I understand them and you tell me if this is how I should proceed. Please note that you did not mention setting the DPI slider. I had thought that the SCREEN data field was LPI and the slider control below it managed DPI:

PreScan 6X6 Transparancy at 48 Bit Color and set crop to full frame.
Set B&W points and Adjust histogram if necessary.
Leave INPUT fields at what crop frame set.
Leave Scale% where it is.
Insert desired lenth and width of final print in inches in the OUTPUT data fields.
Set Q Factor to 1
Set SCREEN to 360
How should the DPI Slider be set or has it been adjsted by the software?
Make and save final scan.

Is this correct ? ( if user input is not required for the DPI slider )

Please see attached FRAME screenshot with settings which were suggested by a friend. How would this work with the 9000 as an attempt to scan at the absolute maximum resolution for the 9000 without interpolation?

Thanks

Re: Scan image image pixel dimension too small?

PostPosted: Thu Aug 06, 2009 9:32 pm
by hassiman
Dear Nils,

On the attached Screenshot The question I was asking about feasability would be using the setup shown but with both SCREEN and DPI fields set to 4000 which is the optical resolution of the 9000.

I realize that the files will be huge... but if this gives me good enough quality for 30 X 30 prints I can always down-rez when making smaller prints... correct? Is 4000 higher than the actual optical resolution of the 9000? If both fields are set to 3600 would it scan without interpolation... or does interpolation only occur when the scan resolution is higher than the rated resolution of the scanner?

Re: Scan image image pixel dimension too small?

PostPosted: Fri Aug 07, 2009 7:55 am
by LSI_Heidorn
Dear hassiman,
there are endless approaches, lets focus on 2 of them:
my approach showed you how to scan perfectly for a given output physical size (e.g. 17" x 17") and a desired Quality ( e.g. photorealistic = 300 dpi ).
In that case we only care for the lpi = Output Res, the input res will be set accordingly to the screen and zoom / Output size, so you do not have to care for that !

The second approach you just mentioned is to always scan in the optical res of the scanner, 4000 dpi would be correct for your scanner.
This obviously leads to huge files and can take a long time but obviously is the most flexible solution.


best regards,

Nils Heidorn

Re: Scan image image pixel dimension too small?

PostPosted: Fri Aug 07, 2009 2:50 pm
by hassiman
Dear Nils... Now I'm betting somewhere... Please clarify 2 last thinhs if you please:

PreScan 6X6 Transparancy at 48 Bit Color and set crop to full frame.
Set B&W points and Adjust histogram if necessary.
Leave INPUT fields at what crop frame set.
Leave Scale% where it is.
Insert desired lenth and width of final print in inches in the OUTPUT data fields.
Set Q Factor to 1
Set SCREEN to 360
How should the DPI Slider be set or has it been adjsted by the software?
Make and save final scan.

Is the set of steps above correct.. and in particular when using your method of scanning for a specific size print does the DPI slider value determined by the software or by the user?



When scanning for the optical resolution of the scanner which value do you recommend be set when scanning with the 9000 to avoid interpolation? 3200 - 3600 - 4000 ??

Thanks,

Re: Scan image image pixel dimension too small?

PostPosted: Fri Aug 07, 2009 4:20 pm
by LSI_Heidorn
Dear hassiman,
here you go :-)

>How should the DPI Slider be set or has it been adjsted by the software?
It will be adjusted by the Software...

>When scanning for the optical resolution of the scanner which value do you recommend be set when scanning with the 9000 to
>avoid interpolation? 3200 - 3600 - 4000 ??

4000 would be the native Resolution of the 9000 !

best regards & have a nice weekend,

Nils Heidorn