Luminous Landscape Forum

Raw & Post Processing, Printing => Printing: Printers, Papers and Inks => Topic started by: Jonathan Cross on February 03, 2009, 04:30:34 am

Title: Dpi for printing
Post by: Jonathan Cross on February 03, 2009, 04:30:34 am
At the risk of rehashing an old topic, can anyone help me?  

I am sometimes asked about the dpi sizing of images for printing, and I trot out '300dpi or thereabouts is a good idea'.  I can be asked from where this figure comes.  I know it is related to the ability of the eye to resolve line pairs, but know no more, despite looking in my old College optics books and doing an internet search.  Can anyone tell me where to find the actual scientific reasoning behind this, preferably in English!  For example, if it is the case that the eye can resolve 6 line pairs per mm, who said so, and when, and under what conditions?

Title: Dpi for printing
Post by: joergen geerds on February 03, 2009, 09:15:36 am
the 300 ppi has a long history.
I think the most common reason was that 152 lpi was the most common screen for offset printing in the 80s and 90s, a kinda middle of the road print quality between news print and high-end magazines. the basic theory said that one should have 2x LPI as a source resolution for good results, or 304 PPI.... most designers can't do any math, so they rounded it down to 300 PPI, because it didn't make any difference in the RIPs (raster image processors that convert the image data into highrez B&W films with screens for printing).

when the first lightjets and similar photo printers came to market, they settled for 3 different resolutions, 200, 300 and 400 ppi, which are the physical resolutions those machines are capable of. and since the printing process doesn't have any screening process, it is considered good practice to feed data to the RIP that doesn't require any resizing (mostly due to volume&speed). recent developments now have "high resolution" digital c-printers with 450 ppi, but only for small prints, not for super large prints.

inkjet/pigment printers use a stochastic screening process, and like offset printing don't require precise resolutions.

but yes, 300 is a good number.
Title: Dpi for printing
Post by: Doombrain on February 03, 2009, 04:12:22 pm
If it's an epson stick to 360, 240 or 180 at the lowest at most printheads have a 180dpi head. the 11880 and new 7/9900 have a 360dpi head.
Title: Dpi for printing
Post by: Wayne Fox on February 03, 2009, 10:06:48 pm
Quote from: Doombrain
If it's an epson stick to 360, 240 or 180 at the lowest at most printheads have a 180dpi head. the 11880 and new 7/9900 have a 360dpi head.

Why does the density of the nozzle itself have anything to do with it? I believe all Epson printers are capable of full 2880 by 1440 DPI.  It just means those with lower density nozzles have to do more passes.  The denser head is one reason the new printers are substantially faster than the older ones .. they can lay down twice as many dots in a single pass.

The relevant "DPI" is how the printer driver handles the data ... which for Epson appears to be 720/360, and for Canon 600/300.

Since any given actual dot can only be one of a few colors, the screening process is far more complicated than just sending your pixels to the surface of the paper.  Personally I find I get terrific results if I send the native resolution to the printer and let the printer driver handle all of the sizing and screening together.  Certainly simplifies the workflow.  Side by side comparisons with prints I have interpolated in photoshop to the magic "360" number are virtually identical with those that I just send at native resolution, as long as I stay above 170-180 ppi.

(Of course this really isn't the OP's question ... think he was really asking how did 300dpi sort of end up being a standard of some type for minimum resolution, which I have read about at some point in the past, but cannot pull the answer nor the source to that out of my aging brain cells.)
Title: Dpi for printing
Post by: TylerB on February 03, 2009, 10:41:36 pm
I think the 300dpi rule of thumb came about for files going to prepress, that's my memory of it anyway. For anyone wanting to get into the nuts and bolts of the issue, this page and the test files are revealing, as Ernst Dinkla has repeatedly pointed out-
http://www.ddisoftware.com/qimage/quality/ (http://www.ddisoftware.com/qimage/quality/)
Downloading the test files and printing at different resolutions and resizing and scaling methods can be very revealing. My experience is similar to Waynes, I let my RIP do the scaling, and a lot of testing has not revealed a visually superior method the vast majority of the time. But rules of thumb do a disservice, and image content file characteristcs can always provide exceptions. Qimages test files and info can show potential problems even when images don't, so there is more to this issue than commonly discussed.
If the question regards upresing small files beyond their real capability, 3rd party tools and photoshops tricks and methods may yield more pleasing results than a straight driver scaling. But in this situation it's the lipstick/pig scenario anyway, as they say...
Tyler
Title: Dpi for printing
Post by: Farmer on February 04, 2009, 02:25:30 am
Quote from: Wayne Fox
Why does the density of the nozzle itself have anything to do with it? I believe all Epson printers are capable of full 2880 by 1440 DPI.  It just means those with lower density nozzles have to do more passes.  The denser head is one reason the new printers are substantially faster than the older ones .. they can lay down twice as many dots in a single pass.

The relevant "DPI" is how the printer driver handles the data ... which for Epson appears to be 720/360, and for Canon 600/300.

Yup.  In a word, yup :-)  I am often glad for your voice of reason and straight-talking, Wayne!

The density of the nozzles has very little to do with it.  The Epsons, for example, use variable dot sizes and extremely complex LUTs in addition to various halftone processing to determine an effective dot pattern to lay down.  The pro level devices can achieve a matrix of 2880x1440 as Wayne says and the consumer level devices can do 5760x1440 (though I dare anyone to pick the differences from 2880x1440 on the same device).

Quote from: Wayne Fox
Since any given actual dot can only be one of a few colors, the screening process is far more complicated than just sending your pixels to the surface of the paper.  Personally I find I get terrific results if I send the native resolution to the printer and let the printer driver handle all of the sizing and screening together.  Certainly simplifies the workflow.  Side by side comparisons with prints I have interpolated in photoshop to the magic "360" number are virtually identical with those that I just send at native resolution, as long as I stay above 170-180 ppi.

Exactly right again.  There is no direct correlation between your image pixels and the individual dots laid down by the printer.  It is only a combination of various dots, of various sizes, in various positions, viewed relative to and in combination with, other dots on variable substrates that provides the illusion of colour (is that a redundancy? ;p ) at a given point.

If you resample (particularly upres) you create data that does not exist to fill the gaps.  The printer then attempts to render this non-existent data as accurately as it can.  Whether that will provide a better or worse result than the printer "filling the gap" itself if you had not upressed will simply "depend".  Sometimes it will, sometimes it won't, most times you wouldn't pick it except in a direct comparison (and not even then many times).

It depends on all the factors involved, starting with the original image all the way through to the final physical printing and the inks and substrates involved and their method of deployment etc etc.

If you have a scene with very sharp, diagonal lines in high contrast to their surrounds, then you're far more likely to see the advantge of higher resolution images - in fact, in that case, if oyu had the data I'd turn on Finest Detail in an Epson driver and send it 720 data.  But in most other cases, you have to ask, "Do I want the printer to do its best rendition of real and fake pixels or do I just want it to do its best rendition of real pixels, even though there are less"?  And only doing test prints will really tell you.

I have one image, that is of the pages of a book, torn and tattered.  The crop from the original 6MP isn't that large so it's a relatively low res image.  Upressing and printing even at A4 shows pixelation from the upressing because the text on the pages is a sticking point.  Printing natively at the same size renders a far more acceptable and desirable image even though it's softer.  Yes, it's hiding a lacking of resolution in the softness, but it's still a better result than a sharp image clearly showing fake pixels.

Quote from: Wayne Fox
(Of course this really isn't the OP's question ... think he was really asking how did 300dpi sort of end up being a standard of some type for minimum resolution, which I have read about at some point in the past, but cannot pull the answer nor the source to that out of my aging brain cells.)

It relates to LPI numbers from pre-press, physical capability of earlier printing devices such as laser printers, an easy number to use and it being over the normal level of human vision to see line pairs (so images looked "solid").  All in all, it remains a good number, but it's not a holy grail and often times effort, time, money and quality are lost in chasing it.
Title: Dpi for printing
Post by: neil snape on February 04, 2009, 02:51:59 am
For press

300 DPI is a line screen often used for ripping to a press frequency of 150 or 152 LPI or DPI if you prefer. The function of the rip is to make a downsampling averaging the information across multiple pixel values to plot a best guess at smooth lines, transitions, etc.

For inkjets

300 DPI falsely put rather being 300 PPI is a number used for setting the grid resolution for driver interpolation into a native format. The interpolation is done resolution wise upres or downsampling to the native operating screen resolution for the driver (or ASIC) performing the most important functions of ripping the files such as masking. The masking resolution and bit depths are the user control settings where in they are toggled for different precisions in the driver settings.
HP's are running mostly at 600 and 1200 ppi sampling from the source grid which being 300 will be a very good compromise. If you send 600 ppi source res, the ripping has a nul transform/conversion if things like max DPI are not user selected. IF so they are usually upressed whatever the source to 1200 for LFP, and 2400 for desktop printers. Inkjet printers perform as frequency modulation printers and or hybrid (Epson variable dot) so the upressing is done for better screening potential, not for diagonal lines as in the case of press AM dots.

Epson seem to run best at integers of 180, Canon and HP run at 300ppi. IF you test your res on MTF charts, you can find the sweet points. What to watch for in images with higher resolution is smoother flat colour zones such as grey areas, and better less fringe coloured high contrast edges. Smoothnes and colour saturation are possible with the printers capable of 5760 or 4800 ppi printing but you need good eyes to find it. Those days are over for me!
Title: Dpi for printing
Post by: Ernst Dinkla on February 04, 2009, 04:26:08 am
Quote from: Jonathan Cross
At the risk of rehashing an old topic, ..............?

That's what I thought in the first place. Then I asked myself why this question repeats itself. The practical value is limited; printer systems exist and it will be hard to do it otherwise with a DIY printer and driver. If there's a want to know for theoretical sake the internet provides good places to get informed.
A google does the job. For example:
http://www.clarkvision.com/imagedetail/eye-resolution.html (http://www.clarkvision.com/imagedetail/eye-resolution.html)
http://www.normankoren.com/Tutorials/MTF.html (http://www.normankoren.com/Tutorials/MTF.html)

If it is about getting a print as good as possible from the image data you have: use a pro printer, pro media and a workflow that has good extrapolation algorithms for up and down sampling. Starting from quality pixels is a must too, enough 600MB MF film scans with void pixels as a result of noise in the film, scanner + inflated scanner resolution numbers.

In practice a shop has to work with any file it gets, the print size resolutions vary between 50 PPI and 1000 PPI. Depending on the media you print on, a range between 150 to 500 PPI could show harder to detect quality differences. With good extrapolation algorithms this becomes even more difficult. With your nose on the print and at a meter distance. So much for eye resolution.


Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/ (http://groups.yahoo.com/group/Wide_Inkjet_Printers/)
Title: Dpi for printing
Post by: Jonathan Cross on February 04, 2009, 07:27:25 am
Thanks for all these replies.  I asked the question as I was originally a physicist,  and like to know why things are as they are, and where science and technology quotes come from.  The web links in Ernst's reply have given me a good start.  Obviously the science has to be modified by the ability of the printer to lay down ink, and by the properties of the paper as well as the ability of the eye to resolve.  The replies clearly show this.  I am much wiser now!

Jonathan Cross

Title: Dpi for printing
Post by: Doombrain on February 04, 2009, 12:01:46 pm
Quote from: Wayne Fox
Why does the density of the nozzle itself have anything to do with it? I believe all Epson printers are capable of full 2880 by 1440 DPI.  It just means those with lower density nozzles have to do more passes.  The denser head is one reason the new printers are substantially faster than the older ones .. they can lay down twice as many dots in a single pass.

The relevant "DPI" is how the printer driver handles the data ... which for Epson appears to be 720/360, and for Canon 600/300.

Since any given actual dot can only be one of a few colors, the screening process is far more complicated than just sending your pixels to the surface of the paper.  Personally I find I get terrific results if I send the native resolution to the printer and let the printer driver handle all of the sizing and screening together.  Certainly simplifies the workflow.  Side by side comparisons with prints I have interpolated in photoshop to the magic "360" number are virtually identical with those that I just send at native resolution, as long as I stay above 170-180 ppi.

(Of course this really isn't the OP's question ... think he was really asking how did 300dpi sort of end up being a standard of some type for minimum resolution, which I have read about at some point in the past, but cannot pull the answer nor the source to that out of my aging brain cells.)

Because the Epson HTM was built on a FM screen of 360 which was designed to accommodate Epson heads.
Title: Dpi for printing
Post by: Wayne Fox on February 04, 2009, 11:02:40 pm
Quote from: Doombrain
Because the Epson HTM was built on a FM screen of 360 which was designed to accommodate Epson heads.

So I guess my question remains the same, what does this have to do with the density of the nozzles, which was the point in your original post.

The lower density of nozzles in the head just means that the older printers must move the media and make another pass to accomplish the same thing the newer printers can do in a single pass.  The literal DPI of the both printers is the same, maxing at 2880 x1440 dpi and has no bearing on the printers screening methodology. Epson could build a printer with a nozzle density of only 60 and produce identical results to the 7900 ... it would just take 6 times as many passes across the media.

Title: Dpi for printing
Post by: mrkahn on February 04, 2009, 11:41:17 pm
I asked this question at the Epson Print Acadamy.  The answer I got was to do as little to change the base resolution of the file as possible.  Iuse a 1DS3 and it produces a 240 DPI file.  The epson said that resolution was more than good enough and better than up sizing to 360 that I thought Epson required for optimal results.  They also said that one would be better off to let the pixels fall where they may as long as they don't get lower than 180 DPI.  I have used this new approach on my new 9900 and it works great.

Just my 2 cents.

Regards,
Malcolm
Title: Dpi for printing
Post by: Ernst Dinkla on February 05, 2009, 04:00:13 am
Quote from: mrkahn
I asked this question at the Epson Print Acadamy.  The answer I got was to do as little to change the base resolution of the file as possible.  Iuse a 1DS3 and it produces a 240 DPI file.  The epson said that resolution was more than good enough and better than up sizing to 360 that I thought Epson required for optimal results.  They also said that one would be better off to let the pixels fall where they may as long as they don't get lower than 180 DPI.  I have used this new approach on my new 9900 and it works great.

Just my 2 cents.

Regards,
Malcolm

Your 1DS3 produces a PPI file and the 240 number goes along with a certain image size in inches.

http://www.steves-digicams.com/techcorner/January_2005.html (http://www.steves-digicams.com/techcorner/January_2005.html)

There are very good reasons not to resample an image file that will be archived for later use. In every workflow though there's a stage where the image file to be printed will be resampled to the input (native) PPI resolution requested by the driver. The driver can do that resampling but some applications you print from can do the same, they intercept calls from the driver (API Windows for example) and do the task of resampling to one of the usual 300-600 or 360-720 PPI native resolutions, sometimes even higher. All on the fly and not changing the image file itself. Depending on the respective extrapolation algorithms the different work flows can show different qualities in print. In the past driver extrapolation algorithms were usually bad but they have improved in time. There still are applications that can do a better job, some even on the fly just before printing . A neglected part in extrapolation is the anti-aliasing for down sampling, I have not seen drivers that cope with that. I'm using Qimage which offers all the elements discussed here in a fast and flexible way.

The rule that you can throw any file resolution in the printing work flow is too general, much depends on the application and/or the driver qualities.

More by Mike Chaney, developer of Qimage:

http://www.steves-digicams.com/techcorner/oct_2008.html (http://www.steves-digicams.com/techcorner/oct_2008.html)
http://www.steves-digicams.com/techcorner/january_2006.html (http://www.steves-digicams.com/techcorner/january_2006.html)

By Bart van der Wolf, on aliasing:

http://www.xs4all.nl/~bvdwolf/main/foto/do...le/example1.htm (http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1.htm)
http://www.xs4all.nl/~bvdwolf/main/foto/do...down_sample.htm (http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sample.htm)


Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/ (http://groups.yahoo.com/group/Wide_Inkjet_Printers/)
Title: Dpi for printing
Post by: Farmer on February 05, 2009, 05:31:15 am
And I still say, do the print and then compare.  There's a lot of talk by interested parties, but the proof (pun intended) is in the printing.
Title: Dpi for printing
Post by: Doombrain on February 05, 2009, 05:57:04 am
Quote from: Wayne Fox
So I guess my question remains the same, what does this have to do with the density of the nozzles, which was the point in your original post.

The lower density of nozzles in the head just means that the older printers must move the media and make another pass to accomplish the same thing the newer printers can do in a single pass.  The literal DPI of the both printers is the same, maxing at 2880 x1440 dpi and has no bearing on the printers screening methodology. Epson could build a printer with a nozzle density of only 60 and produce identical results to the 7900 ... it would just take 6 times as many passes across the media.

yep, the 7/9600 range has heads with only 90dpi, then the x800/880 with 180 and now 360. the speed isn't doubled, it's more like x1.6 quicker.
in terms of screen epson devolved the screen algorithm based on a 90 x 6 nozzle head (for photo) so the driver will always interpolate data to 360 or 720 if you check finest detail. The nozzle pitch is everything when producing the screen.
Title: Dpi for printing
Post by: Farmer on February 05, 2009, 06:05:27 am
Quote from: Doombrain
yep, the 7/9600 range has heads with only 90dpi, then the x800/880 with 180 and now 360. the speed isn't doubled, it's more like x1.6 quicker.
in terms of screen epson devolved the screen algorithm based on a 90 x 6 nozzle head (for photo) so the driver will always interpolate data to 360 or 720 if you check finest detail. The nozzle pitch is everything when producing the screen.

The heads aren't 90/180/360 dpi - the current TFP heads are 360 nozzles per colour channel, it's nothing to do with inches in that sense of DPI and the driver has been completely redesigned around new LUT technology.

And, yes, 360 and 720 remain important numbers, but I'll say this again - print some tests before you assume that creating pixels is better than the driver figuring it out.  To make a blanket statement one way or the other is, frankly, wrong.
Title: Dpi for printing
Post by: Doombrain on February 05, 2009, 06:31:27 am
Quote from: Farmer
The heads aren't 90/180/360 dpi - the current TFP heads are 360 nozzles per colour channel, it's nothing to do with inches in that sense of DPI and the driver has been completely redesigned around new LUT technology.

And, yes, 360 and 720 remain important numbers, but I'll say this again - print some tests before you assume that creating pixels is better than the driver figuring it out.  To make a blanket statement one way or the other is, frankly, wrong.

The LUT only determines the colour output combination, not the screen.
The FM screen (HTM) was updated when the Pro3800 came out, until then it remain the same for years. The update was to improve transition curves to make better use of VDS 1 to 3.

For the most part is doesn't really matter what PPI you use as long as you don't go below 180. If you use something like 369PPI there's a chance of introducing Moiré patterns.

I wasn't making a blanket statement, I was just responding to a question regarding best output PPI and i answered the question for Epson printers, which i know alot about. Just trying to help out.

I don't want to become the villain here, it's obvious you're all very passionate about PPI. I've said my bit.
Title: Dpi for printing
Post by: Farmer on February 05, 2009, 06:46:40 am
I wasn't really directing anything at you :-)  I agree totally that you don't need to worry so much about the PPI put through.
Title: Dpi for printing
Post by: Charles Gast on February 05, 2009, 11:33:32 am
In the tutorial Camera to Print Michael and Jeff strongly assert the advantage of printing with "native resolution". That means don't upsample or downsample. The engineers who designed the z3100 say the same thing. Just send it 180 to 400 dpi and viola!  Jeff also seemingly suggests that genuine fractals is snake oil = )    Here's the catch:  **The Output**  In most prints native resolution is ok, But I printed a photo of a cat. When I sent the z3100 an (approximately) 204 dpi image with zero up/downsampling the cat has several ziggy zaggy whiskers... The jaggies stand out in the print quite visibly with no need for a magnifying glass or even my reading glasses. The jaggies are not there on the monitor. I went back to the pre-sharpened image and used the snake oil application (genuine fractals) to upsample to 300 dpi, used photokit sharpener with a little extra sharpening following the upsample and printed. The result was sharp clear cat whiskers free of jaggies on the print. So according to what I see on prints many if not most prints are ok with native resolution but if there is very sharp contrasty detail it may at least at times be better to go with the old 300dpi to the hp.  The print driver does some resampling and I guess under certain circumstances it is best to send it a file with which it provides more desirable output.
Title: Dpi for printing
Post by: Tklimek on February 05, 2009, 12:44:05 pm
Charles....

In the Lightroom 2 tutorial, Jeff has indicated that in SOME images (I believe high-frequency) testing has shown to produce better printed output by uprezzing.  Check out the LR 2 tutorial and see....did anyone else catch this in that video?

Todd in Chicago

Quote from: Charles Gast
In the tutorial Camera to Print Michael and Jeff strongly assert the advantage of printing with "native resolution". That means don't upsample or downsample. The engineers who designed the z3100 say the same thing. Just send it 180 to 400 dpi and viola!  Jeff also seemingly suggests that genuine fractals is snake oil = )    Here's the catch:  **The Output**  In most prints native resolution is ok, But I printed a photo of a cat. When I sent the z3100 an (approximately) 204 dpi image with zero up/downsampling the cat has several ziggy zaggy whiskers... The jaggies stand out in the print quite visibly with no need for a magnifying glass or even my reading glasses. The jaggies are not there on the monitor. I went back to the pre-sharpened image and used the snake oil application (genuine fractals) to upsample to 300 dpi, used photokit sharpener with a little extra sharpening following the upsample and printed. The result was sharp clear cat whiskers free of jaggies on the print. So according to what I see on prints many if not most prints are ok with native resolution but if there is very sharp contrasty detail it may at least at times be better to go with the old 300dpi to the hp.  The print driver does some resampling and I guess under certain circumstances it is best to send it a file with which it provides more desirable output.
Title: Dpi for printing
Post by: TylerB on February 05, 2009, 01:01:50 pm
If you simply run the tests for yourself, rather than pondering how it should work based on numbers and characteristics of drivers only partially known to us, or advice put forth by supposed experts putting forth ideas from other experts, you will get to the bottom of this issue. Mike Chaney's tests are revealing, they proved my previous preconceptions false.
How this relates to optimizing a workflow depends on a number of factors. Landing on hard rules that are absolutely optimal for every file and print is a goal riddled with problems, in life, art, and craft.
If the interest is primarily in the hard science, there is too much we don't know about the various driver's scaling. We get these native res numbers, which are somewhat illuminating, but clearly black box driver or RIP scaling methods have been evolving.

Tyler
Title: Dpi for printing
Post by: Doombrain on February 09, 2009, 06:59:50 am
Here's a quick test for the people who think print output PPI isn't that important. Epson Pro3800 (doesn't matter which Epson you use) and Epson driver.

Top image is a RAW file from a R-D1 processed at 360PPI and printed at 720x1440
Bottom image is the same only processed at 300PPI.



Edit: make sure you view at 100%
Title: Dpi for printing
Post by: neil snape on February 09, 2009, 03:11:17 pm
Yup the above is just one example of the differences seen with varying resolutions.

What you will find is if you print at higher resolution , let's say 360 then source at 720 with maximum printer res, tht gray repeating contrast lines will have less coloration, and less staircasing. Solid flat areas will have less , if not much less mottling.

Quite a few things change , but it takes many tests to actually pinpoint cases where it makes all the difference as shown in the example above.

Title: Dpi for printing
Post by: Farmer on February 09, 2009, 05:32:06 pm
Quote from: Doombrain
Here's a quick test for the people who think print output PPI isn't that important. Epson Pro3800 (doesn't matter which Epson you use) and Epson driver.

Top image is a RAW file from a R-D1 processed at 360PPI and printed at 720x1440
Bottom image is the same only processed at 300PPI.



Edit: make sure you view at 100%

When you say "processed" what do you mean?  What is the native resolution of the image file?  Did you resample to achieve your 360 and 300 output at that partcular print size?  Or did you print at a size that meant that native was 360 and then you downsampled the file to print at the same physical size at 300 or was the native 300 and then you upressed to 360?

If you resampled the file, what method did you use?  Did you try 720x720 and 1440x1440.  Did you print bi-di or uni-di?

Also, I assume you're saying that the pattern seen in the 300 sample has scanned accurately and is the same as visible in person?  What settings did you use when scanning?

Yes, there are most certainly times (as I think most people have stated) when choosing 360 (or even better, 720) with an Epson driver will help, but in most cases, most people won't pick the differences and in some cases upressing will give you a poorer result.

Did you also try at 240 and 180?  And what about 313 or 297 or something else odd?  If you want to make a sweeping statement that upressing to 360 is always better, then you really need a bit more data and to explain your methodology :-)
Title: Dpi for printing
Post by: Doombrain on February 10, 2009, 03:59:27 am
We seem to be running in circles here, this isn't a sweeping statement, it's a fact.

My point is the native res of the print Epson driver is 360 so if you stick to integers of 360 you'll be OK as long as you don't go below 180 to avoid pixelation.

Native Image Res;
The two files are camera RAW files opening in Adobe Camera RAW.
The first one was set to 360PPI then opening in photoshop for printing.
The second image was set to 300PPI and then opened and printed on the same sheet.

The file was printed at 720x1440DPI, but it doesn't matter which DPI you set as long as it's not a draft mode (you might even see it at a draft mode).
Mode was Finest Detail Off, High Speed On.

To scan in I used an integer of the scanners native res, which was 96PPI to capture the effect. The scan is off a single sheet, no trickery.
Title: Dpi for printing
Post by: Farmer on February 10, 2009, 05:03:21 am
Quote from: Doombrain
My point is the native res of the print Epson driver is 360 so if you stick to integers of 360 you'll be OK as long as you don't go below 180 to avoid pixelation.

You can certainly print below 180 without pixelation depending on the image and the substrate.  Yes, you need to be careful, but really 120 is the point at which you really need to be concerned in most cases.  Obviously it won't be suitable for some work, but you can certainly get down that far very often.

Quote from: Doombrain
Native Image Res;
The two files are camera RAW files opening in Adobe Camera RAW.
The first one was set to 360PPI then opening in photoshop for printing.
The second image was set to 300PPI and then opened and printed on the same sheet.

OK, so they were both printed to the same physical size?  That doesn't tell me if either of them is the native resolution in print.  If the image is 3000x2000 pixels and you printed it on an A4 with a border you might have printed it at say 11"x7 1/3" then it's printing natively at about 272, so at either 360 or 300, it's being interpolated by Photoshop before being printed, which isn't the query.

The query is whether upressing or down sampling to 360 is better than printing it without any resampling.

Now it's entirely possible (likely!) that one your resses was native for the size being printed, but you haven't confirmed which.  At the size printed, was it 300 native and you upressed to 360?  That's valid.  If both were simply opened at 360 and 300 respectively and then printed without resizing then they would print to different sizes and you're not answering the question at hand, which is if you want a particular size should you resample to 360 or go with what you have?

Quote from: Doombrain
The file was printed at 720x1440DPI, but it doesn't matter which DPI you set as long as it's not a draft mode (you might even see it at a draft mode).
Mode was Finest Detail Off, High Speed On.

It does matter.  It absolutely affects the dot pattern being laid down and that can affect the output.  It's absolutely valid to test at 720x1440 is that's the resolution you want to use, but it's not the most common resolution to choose on an Epson Pro printer, so it's worth checking against other options.

Quote from: Doombrain
To scan in I used an integer of the scanners native res, which was 96PPI to capture the effect. The scan is off a single sheet, no trickery.

I would recommend only scanning at the scanner's native physical resolution in order to avoid any errors being introduced.  I wasn't for a moment suggesting any attempt at trickery!  96ppi is really too low. imho.

I don't think we going in circles.  You have valid points and a worthwhile test to pursue, but there are things that either need to be confirmed or explained or, perhaps changed in order to validate it.

As Neil said, you can make a test to obtain any result that you want.  If I have time tomorrow at work, I'll find an image and print it at various "native" resolutions and then at corresponding resampled resolutions - to be honest it may take a day or two depending on how busy I am.  I like your approach of showing some actual examples for people.
Title: Dpi for printing
Post by: Ernst Dinkla on February 10, 2009, 05:08:27 am

Nobody takes pictures with the 300 PPI or 360 PPI of resp HP, Canon or Epson printers in mind. Nobody should either.
The moment the print size is set the only thing that counts is getting a workflow with the best extrapolation aboard. There's no way to avoid up or downsampling to either 300-600 PPI or 360-720 PPI as that is part of the system, so do it as good as possible. After that the print quality is the result of the printer quality, the media coating, the extrapolation quality, the original image data. All things equal but the extrapolation algorithms varying will show print quality differences if a certain general quality in the other components is available. To check the print quality of your existing workflow use one of ddisoftware's targets and its test method.

If one describes the print quality difference between 300 PPI and 360 PPI input to a 360/720 PPI system, both input files based on the same image data, one actually describes a difference in extrapolation algorithm quality and not the superiority of either 300 PPI or 360 PPI input.



Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/ (http://groups.yahoo.com/group/Wide_Inkjet_Printers/)
Title: Dpi for printing
Post by: Doombrain on February 10, 2009, 05:55:24 am
@ Farmer

180PPI is the lowest I would ever go for any of my clients simply because I'm sticking to a rule of multiples.

There's no resampling in Photoshop. Photoshop was only used to open the RAW files, crop tool was used to make the sizes the same and print the images.

IMO the best possible workflow is not to do any resampling and stick to the native res of the print driver, in this case 360PPI. But only if possable.

DPI, it was just a casual passing comment in terms of this test.

Scan res was set at 96PPI because I wanted to upload the image here and it demonstrated my point.
Title: Dpi for printing
Post by: Farmer on February 10, 2009, 06:43:43 am
It's usually not practical to print at different sizes to accomodate natively maintaining a particular DPI.

I printed about 100 images for a major photojournalism awards last year - some were the same pixel dimensions because they were uncropped from the same camera model (despite being from different PJs), but many were different.  All had to be the same physical size output (or at least contained within a certain physical specification, allowing for different aspect ratios).  That's the more commons scenario, printing to a reasonably standard output size.

That being the case, people want to know whether to resample or to leave it "as is" assuming the resolution is within a certain tolerance.  Yes, changing print size to accomodate available resolution will always give you the best results, but it's not a practical real-world option most of the time.  My experience is that resampling the file can often do more harm than good or simply isn't necessary.  There are most definitely situations where it is beneficial, but it's not some holy grail.

I'll find time during the week and do some samples and tests if I can.

BTW, 120 is an "integer" of 360.
Title: Dpi for printing
Post by: Farmer on February 11, 2009, 02:15:49 am
Well I had a chance today to run off some prints and scan them in.

All were done on an Epson Pro 11880 at 720x1440dpi super microweave on, bi-directional, no colour management (colour managed from Photoshop) on Epson Premium Glossy Photo Paper 170 using the Epson canned profile with one exception - I made one print from Qimage.

Of a total of 7 prints, 4 were printed to the same physical size (about 16"x9") and 3 were printed physically smaller by using the native pixels at higher resolutions as I'll detail below.

The image is 2076x1168 pixels (a crop of the full frame), opened in ACR 5 where adjustments were done, including sharpening for print at medium level.  For Qimage a PSD was saved out and then printed.  At the chosen print size, the native pixels provided 129ppi resolution.

The following prints were done:

Upressed to 360ppi in PS CS4 standard
Upressed to 360ppi in PS CS4 smoother
Upressed to 360ppi in Qimage as a result of printing to the specified size
Printed at 129ppi native in PS CS4
Printed at 360ppi native in PS CS4
Printed at 240ppi native in PS CS4
Printed at 180ppi native in PS CS4

The last three obviously printed smaller than the approximate 16"x9"

All of the full size prints were scanned in using an Epson V700 at 300dpi with unsharp mask off and descreening off.  The prints that were smaller were scanned at appropriately higher resolutions to provide the same pixel output to the saved scan.  The scans are of just a portion of the prints to avoid the files being too large.

The 7 images are linked here for those who are interested, in no particular order (ie does not match the order above):

Image A (http://www.the-farm.net/LL/a.jpg)

Image B (http://www.the-farm.net/LL/b.jpg)

Image C (http://www.the-farm.net/LL/c.jpg)

Image D (http://www.the-farm.net/LL/d.jpg)

Image E (http://www.the-farm.net/LL/e.jpg)

Image F (http://www.the-farm.net/LL/f.jpg)

Image G (http://www.the-farm.net/LL/g.jpg)

In the next post, I'll list which are which.

The scans don't tell the full story, as the real test is how it looks to direct observers and not a resampled (ie scanned) copy of the print.  I hope some will find it of interest/use to compare.  It's a not a definitive "do it this way" or "do it that way" and I'll let the reader determine their own conclusions (my preferences are well posted :-)
Title: Dpi for printing
Post by: Farmer on February 11, 2009, 02:18:42 am
A - 129ppi native

B - 180ppi native

C - 360ppi Qimage

D - 360ppi upres PS smoother

E - 360ppi upres PS normal

F - 360ppi native

G - 240ppi native
Title: Dpi for printing
Post by: Ernst Dinkla on February 11, 2009, 07:06:45 am
Quote from: Farmer
A - 129ppi native

B - 180ppi native

C - 360ppi Qimage

D - 360ppi upres PS smoother

E - 360ppi upres PS normal

F - 360ppi native

G - 240ppi native

The subtitle of this thread is "What's the science ?". While we are wandering off the original context of this thread it may be a good idea to keep some quality in the discussion. It will have taken you some time to do the job but for me it isn't telling me anything. Comparing image quality of different sized prints on scans made of the prints is already beyond science.  On top of that there will be an extrapolation step in the scanner driver again with the method you used. It also looks like color management between Qimage and Photoshop isn't synchronised or the scanner fluctuates in its color management.

The use of the term "native" gets confusing too, for some it means the resolution of the original image but then it only has meaning if the size of the image is included too, for others it means the resolution required by the printer driver for certain media settings (= print quality) in the driver. As written before the last varies for different printers and printer brands. Today that's more or less 360-720 (-1440) PPI and 300-600 (-1200) PPI, the last high number for printing of vector designs and text. Some RIPs are able to use different numbers like 450 PPI on my Wasatch RIP for an Epson 10000. I don't think it is wise to use the term "native" resolution for image data, it is as flexible as the size is (even without extrapolation).

If there are samples to be compared it is C, D and E while Qimage actually knows a lot more of extrapolation routines that may suit the subject better. Smart print sharpening considered? Compare the prints at eyesight with a lot of diffuse light or if there has to be a representation here, scan them all on the V700 at 1800 PPI, way beyond the print resolution and reducing the influence of extrapolation in the scanner driver. Small image crops may show area artefacts of the paper coating so will not tell what the general quality difference is. I prefer a written comment on the quality difference by someone who can judge prints above scan samples on the web.


Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/ (http://groups.yahoo.com/group/Wide_Inkjet_Printers/)
Title: Dpi for printing
Post by: Farmer on February 11, 2009, 03:38:12 pm
Thank you for the feedback, Ernst and I do agree with you.  In respect of "native" I did make a point of listing the ppi that was sent and explained what was going on, so hopefully people can understand the process in each case.

I also agree that scanning adds variables but shor of sending sample prints to everyone I've little choice :-)  I can tell you that my observation is that there is little difference between those images printed at the same physical size (I indicated 16"x9" as the "full size" prints) and in a quick blind test at work, no one was able to pick between the same sized PS prints (all prints observed in a light booth).  Obviously the Qimage one looked different in this case, but there was no particular preference or otherwise against it in terms of apparent resolution and sharpness.  I might try a Genuine Fractals enlargement and see if tht is any better.

I can certainly scan again at higher resolutions - it just needs to be a sample of a smaller area to avoid massive files.  You also need to be familiar with looking at high-res scans I think to make use of that.

I would encourage those keen to know to run their own tests and observe the results directly, as you're quite right that direct observation is much better than 3rd hand via a scanner.
Title: Dpi for printing
Post by: fike on February 11, 2009, 05:35:14 pm
I don't really think that native resolution means anything.  What are the dimensions in pixels?  That tells you how much data you have to express your image.  From there, I try to keep the printer output resolution above 240 DPI--divide the pixel dimensions by 240 and that will give me the largest dimension in inches.

If you need to modify your sharpening, resizing, whatever to get better results, go for it.  I use Qimage and it is give me very good results as long as I stay above 240 DPI (220 in a pinch).  Beyond that, if I go to 800, 900 1000DPI who cares.  I decide the DPI based upon the size I want the image to be printed, not the dimensions of the image.  

It is almost like saying that you have a 5DMKII with an image output size of 5616 x 3744, therefore if you assume at 360 DPI then its native image size is 15.6" x 10.4".  That is crazy!  The image is whatever size you want it to be.  Resize the crazy image and then sharpen or blur the image as needed to get the best results you can eek out.  You may need to experiment with various preparations based upon the subject.  

There are no hard and fast rules.  There are thresholds that people might discuss like 240 DPI versus 360 DPI as the minimum resolution, but print some images and make up your mind yourself.  Different images, paper, and printers may demand different preparation.  There are lots of variables to mess with, but some notion of native resolution of the image is not relevant.