I'd been accepting the Internet advice - I think I saw it here on LL - that highest print quality results from up-resing 16-bit files to 600 ppi with Bicubic Smoother, rather than letting the printer interpolate a 300 ppi file to 600.

But in experimenting with this, I came up with unexpected results.

I saw on the 5000 Wiki that someone noticed no difference between 6x9 prints from 300 & 600 ppi files. But this raised the question of what happens with larger sizes. I usually print 13.5x20.25 on 17x22 sheets.

First of all, the street smarts were correct: A print from a 600 ppi file res-ed up with PS's Bicubic Smoother did prove to be sharper than one from a 300 ppi file, interpolated to 600 by the printer. In addition to the difference in sharpness, the file interpolated by the printer seemed to have darker & muddier lower midrange/shadows.

But after confirming the received wisdom, I tried res-ing up with Genuine Fractals.

Initially I tried going all the way from the native image size @ 300 ppi to 13.5x20 @ 600 ppi, but GF was too slow for this to be practical - I could literally go out for a quick lunch while my middle-aged G4 struggled through the computations. (Maybe I can work entirely in GF with a new Mac Pro & CS3, later in the year.)

Second, I tried up-resing the 300 ppi image to 13.5x22 in GF, & then letting the printer do the further increase to 600 ppi. Compared to the 2 previous prints, this one was noticeably sharper to the eye (not just through a loupe!). And the lower midrange & shadows were as clear as in the BiSmoother print.

And even though GF is a slow program - especially on 16-bit files - it was faster to work with GF & a 300 ppi file before PhotoKit Output Sharpening than to sharpen a massive 600 ppi file made with BiSmoother. (This size of print balloons to 1.65 GB in the course of sharpening.)

This led me to conclude that both Bicubic Smoother res-up & iPF5000 printer conversion to 600 ppi have disadvantages compared to using a better res-up 'engine' in the first place.

(I used Genuine Fractals because there's no QImage for Macs. For anyone who cares about the whole workflow: 5D file processed in ACR & ProPhoto workspace to 'native' size @ 300 ppi. PhotoKit Capture Sharpening = Hi Res, Narrow Edge; PK layer opacity reduced to 66%. After PS work, res-ed up in the 3 ways described above, followed by PK Output Sharpening = Inkjet 480 Matte; PK layer opacity reduced to 88%. Printed on HPR thru plug-in & cassette.)

Kirk

Where did this data come from? 13.5x20.25 @ 300ppi is 24mp. If you're starting with less real data than this, you won't see what your printer is capable of. Try printing your original at 300ppi with no upsampling. Anything larger and you'll just be trading resolution for size.

Kirk,
Thanks for posting your results. I'm the one who posted on the wiki about testing 300 vs 600ppi images through various driver paths.

In the test that I posted on the wiki I wasn't looking at interpolation, the images were at native resolution. So that was a pretty easy problem even for the driver interpolater. So at about 8x10 the differences were negligible. But my image type certainly plays a factor and, as you found, the amount of interpolation plays a big part. The big news for me was that I didn't need to carry 16bits in the print file, although I keep my masters at 16bit in case I need to do future adjustments.

I had been evaluating Genuine Fractals, QImage, and PS CS2's bicubic smoother for upressing and talking to other photographers in other forums. A technique recommended by another photographer, outlined by Jack Fletcher on Digital Outback Photo, Uprezzing Digital Images (PC & Mac), showed interesting results. There are actions for PC and Mac that implement the technique linked to the page. I had decided that Qimage's hybrid interpolation was as good or better than Genuine Fractals with Qimage being much less expensive, so I bought Qimage SE. I later compared the DOP action above to Qimage's Hybrid SE interpolation on images upressed nearly 400% from approx 8x10 to 16x20. The results were very impressive and nearly identical . In fact I unwittingly printed a 2.5x2.5@300ppi image at 8x10 (1300+%) and it looks pretty good. So if you haven't tried the DOP action I'd recommend getting it.

The results from Qimage, which will auto-upres from native resolution to whatever I tell it to print at, and getting good results printing through the OS driver with a good profile opens up a simpler workflow for me for most images rather than upressing seperately then printing through the plugin. Still, I think there are advantages to using the plugin in how it controls the printer, so when I want the best print I go that route.

I have an idea about how I might be able to drive the PS plugin from Qimage which would be the best of both worlds. Not much help for the Mac guys but if I figure it out I'll let the forum know.


Anyway, I was interested to see your results.

-- Kalin

I am curious, if you looking for sharpness why are you using ACR to process your files. I thought ACR was for people doing portrait or other photos that wanted a soft look. I just find it strange that you go to that much trouble to sharpen or to achieve a sharp image and then use ACR to start with.

Maybe I am missing something and you consider detail and sharpness two different things.

Doyle

Did you use bilinear in the plugin to uprez? It is better than nearest neighbor. My compromise is to uprez the original file to 6"x9" 600ppi using PS bicubic smoother
then sharpen with smart sharpen, lens blur and save as a master file. I then feed the plugin this file and let it uprez to the print size using bilinear. Yes GF and Qimage and upressing all the way up then sharpening is (maybe) a tad better but not worth the time effort and file size. Also what paper are you using for these tests and are you using a custom profile specifically for the plugin?
Marc

The choice of ACR is not a factor here. ACR is definitely not only for doing portraits or soft-look images. ACR, like all of the other good RAW converters, produces sharp images as long as it is used properly. Remember that most RAW images will need some sharpening applied to bring out the edge detail. If you turn off sharpening in the raw converter, then sharpening will have to be applied later.

Eric

For my tests I'm using a custom profile for the plugin and Epson Enhanced Matte and Epson Premium Luster papers. When judging interpolation results I've been upressing the files and comparing them in CS2 at 100% or higher looking at edge sharpness and detail. I haven't printed each of the results in the interest of ink.

I'm been doing my RAW processing in DxO and lightroom beta 4 vice Adobe Camera Raw (ACR).

Marc, after seeing your post Kirk's original post may have been in reference to your 6x9 workflow rather than the resolution test I had posted on the wiki. Doesn't really matter either way, it just occurred to me.

Kalin

I thought the same so I answered this just incase there was a misunderstanding
BTW I used the following for my tests 5D, DXO, PSCS2, custom profile and Fuji Pro (high gloss and very white) over time I just got tired of uprezzing and sharpening for an almost unperceivable improvement-Marc

Thats interesting as I could never get any acceptable detail no matter what settings I tried. Seems any sharpening I tried in ACR or later never really gave more detail just edge definition. I could never get anywhere near the detail from ACR like I can in Nikon Capture or Raw Developer so I just assumed ACR was like most other raw convertors and was design to produce soft images.

Doyle

ACR doesn't produce soft results... The results are only withput sharpning. If you REALLY turn of all sharpening in C1 or other raw convertes, than they are all soft at the end. The diffrence is that from default most other convertes have a much higher value for sharpening. If you use PS to sharpen later, what I would always do. ( don't let any raw converter sharpen your image!) You get nearly the same results from all of them.

Well I have tried you technique and I wasn't impressed. I think I have tried/demoed every converter out there. The only two that I have found that gave great clean (lack of artifacts) detail whether sharpening in the converter or later, are Raw Developer and Nikon Capture/NX.

If you find ACR works fine for you I am certainly not argue with.

Doyle

Now everyone is own. I certainly won't waist any time on doing some test and shwoing them, because what you will get is that 5 diffrent pictures are so close that you won't be able to say which is softer or sharper. Now I never used ACR 2.4 but because of the colour and not because of sharpenes. But I really think that will cange with the new CS 3 versionm, because what I have seen and tested so far it really is the best converter out there. (Pherhaps raw dev. is really close but that thing is only for mac)

I also don't know how you sharpen in PS but that's so much better than any raw converter. You get much better files.