PDA

View Full Version : Output sharpening for large prints at 100 ppi


Christopher_Campbell
03-05-2005, 05:42 PM
I've been using Photokit Sharpener very successfully this past year, and it has always done everything I could have asked. Now, however, I've just purchased a wide format printer (an HP 130nr), and need assistance on an appropriate workflow for enlarging the occasional Canon 20D frame to the full printable size of this printer, which would be almost 24 x 36". After working through Bruce's Real World Camera Raw, processing in ACR at native resolution and applying capture sharpening is no problem, and I've long since developed a feel for analyzing my images and determining the appropriate settings according to subject edge frequency.

I realize that there is a considerable divergence of opinion about interpolating file for large prints (I've read most of the threads on the subject in this forum), but with my Epson 1280, I have tended to follow Bill Atkinson's advice that it is better to use Photoshop to get the data up to the Epson rasterization value of 360 ppi, and so have used Bicubic Smoother to upsample my files when I wanted to print them 13 x 19". On a fast G4, my 16-bit 20D files would bloom from 47 to 166 MB, which is big, but still entirely manageable. Then using Photokit to output sharpen the upsampled files has produced superbly detailed and sharp prints on the 1280 when printed at 2880 dpi.

Thinking about a print size as large as 24 x 36" from an 8 MP camera is obviously a somewhat more difficult problem. The native resolution of the 20D, 2336 x 3504, can only provide just under 100 ppi of actual data at that size. I did a test today in which I upsampled a crop (roughly 1/9th of the image) by 3X at 300 ppi (the rasterization value of the HP 130), which is precisely equivalent to the conditions of upsampling the full frame to 24 x 36". When I then applied Output Sharpening (Inkjet 300 Glossy) to the file, and previewed the effect on the screen, it appeared that the sharpening effect was relatively minimal, probably because the action was intended to be used on actually captured pixels, not pixels manufactured by interpolation. So what to do?

At 100 ppi, I'm way below the lowest Output Sharpening set (Inkjet 180), so it doesn't look as though I should capture sharpen at native resolution and then interpolate up (or not). I printed out the test section, and it looks definitely soft (suprise, surprise!). However, my instinct is that with certain files where I'm using very careful technique ? big tripod, Canon's best primes, mirror lockup, remote release, moderate apertures to balance MTF and diffraction ? there ought to be a way to get a satisfying large print. Looking at the test print section, my feeling is that everything will fall into place once the sharpening is pushed a little harder, but I'm just not sure how to go about it.

Thanks in advance, Christopher Campbell

Bruce Fraser
03-05-2005, 06:06 PM
I suggest uprezzing to 150 ppi, and using the Inkjet 180 Output Sharpener. I don't think more uprezzing will buy you anything. (But I could be wrong, so test, and if you feel inclined, let me know!)

Christopher_Campbell
03-05-2005, 06:15 PM
Bruce, Thanks for the ever-so-quick reply. I will try that, and make another test print. I've just spent a day doing resolution testing of the HP 130 (1200 vs. 2400 dpi) as part of a thread on Luminous-Landscape,

http://tinyurl.com/6m8cr

and so am fully in the mode of examining pairs of prints to distinguish fine differences. I'll report back on what I find.

Christopher_Campbell
03-05-2005, 06:18 PM
Argh, the link above doesn't work because the Luminous-Landscape forums require log in. Sorry! For anyone interested in a thread on the HP 130, see the thread, "Really exciting printing times ahead of us" in the "Printers, Papers and Inks" forum.

Christopher_Campbell
03-09-2005, 08:39 AM
Bruce,

Ever since I purchased Photokit Sharpener, I've had the feeling that some day I would have to really roll up my sleeves and burn some paper to understand exactly what it was capable of doing. Now that I'm on the other side of that project (or at least round 1), I'm ever so glad for every minute: yesterday I finally had gorgeous 23.6 x 35.4" prints rolling out of my HP 130nr printer from my 20D files! I ended up doing something like twenty 8 x 12" swatch prints in a very systematic attempt to compare a variety of techniques. I intend to scan sections of the test prints and post a fairly complete report in the next week or so, but here's the short version. I compared:

1. Process in ACR to native resolution (2336x3504), capture sharpen, upsample to 300 ppi, output sharpen 300.

Result: unacceptably soft

2. Process in ACR to native resolution, capture sharpen, upsample to 150 ppi, output sharpen 180.


Result: slight improvement in sharpness and contrast over #1

3. Process in ACR to native resolution, capture sharpen, upsample to 300 ppi, creative sharpen: edge sharpen, output sharpen 300.

Result: with addition of creative sharpening, slightly sharper than #2

4. Process in ACR to native resolution, capture sharpen, upsample to 150 ppi, creative sharpen: edge sharpen, output sharpen 180

Result: with addition of creative sharpening, significantly sharper and more detailed than #3

5. Process in ACR to highest available resolution (4096x6144), capture sharpen, resize (no resampling!), creative sharpen: edge sharpen, output sharpen 180

Result: much sharper, and much more finely detailed than #4

6. Process in ACR to native resolution, capture sharpen, upsample to 120% @ 300 ppi using Bicubic Smoother, creative sharpen: edge sharpen, downsample using Bicubic Sharper, output sharpen 300.

Result: no advantage; comparable to #3

To my great surprise, #6 is far and away the best workflow in this specific instance. My first conclusion is that pixels generated in ACR's linear to gamma conversion extract significantly more information from the raw capture than pixels interpolated later after a conversion to native file size. Second, when pushing 8 MP to 100 ppi, the additional sharpening of the creative edge effect sets is necessary to convert excessive softness into very fine pattern. With a larger capture, this would probably need to be dialed back, or eliminated. Third, upsampling to match a printer's rasterization value obscures subtle detail instead of enhancing it. Fourth, forget arm-chair theorizing and conventional wisdom. Think, keep scrupulous records, make test prints and move forward! Thanks again.

soboyle
03-09-2005, 10:36 AM
Christopher
Thanks for the post, helpful information, one question, in your conclusion you say that #6 is the best method, but in your detailed notes you seem to indicate that #5 is the best method, just want to clarify that.

Christopher_Campbell
03-09-2005, 10:48 AM
Shaun, You are absolutely right, #5 in my list was the best workflow. I wrote the conclusion, then resequenced the examples to make a more logical sequence, and forgot to correct the conclusion. Thanks for pointing that out.

Steve Terrell
03-09-2005, 12:08 PM
Interesting information Christopher. For what it's worth I have taken a 10D file up to 24x36 upsizing first in ACR as you did and obtained amazing results. I was blown away by the quality of the print. Of course you nedd a properly exposed file to begin with and careful handling, but it can be done.

Steve

Rapt
03-23-2005, 09:37 AM
This is GREAT information. Thanks guys! This is worth the price of admission ... LOL!

My solution (for subjects that allow it) has been to stitch multiple images to give the size and resolution required to print at a full 300-360 ppi. (Of course there's tons of processing time and labour involved in that.)

But sometimes the subject won't allow this of course, or you get an image that you just can't go back and add to. So this is invaluable!

yngve
03-24-2005, 11:36 AM
Cristopher

Your work is really interesting! I am really looking forward to the complete report you've mentioned!
I print on a Epson 2200, and I would love to optimalize my workflow to improve the percieved sharpness in my output. I also have the DxO software, wich dosent allow any upsampling, but since it can output DNG files, I can upsize when I convert in ACR. Disabeling the DxO sharpening off course.

yngve