A new Printer/Media Resolution test target

[Bart_van_der_Wolf]

With the newest TIFFs my odd looking slanted bands moved up from 8 into the 10 area on the right page (In my images posted prior.).  Also, the pure white and black in the upper right is now smoother and gray tones (Not the stark B&W prior.).  So something changed in the new iteration.

I do seem to seem more lines going into the middle of the circular target as well.  Qimage still seems to be the best for even tonality over PS though.  PS CS6 seems slightly more contrasty which leads to an apparent sharpness increase.

Hi SG,

It sounds like something else changed as well on your side, because e.g. the middle region towards the circles (the star and the circles) is unchanged between target versions. Only the numbered patches and frequency sweep were improved a bit (fractionally smoother and more accurate), and it is now a TIFF file (to avoid issues with the puzzling Adobe PPI tag in PNGs). The fact that you moved from 8 into 10 cy/mm is very nice, but may be caused by the same thing that changed the printing of the unchanged center region. Maybe the (Adobe RGB) profile that was embedded in the PNG wasn't picked up for some reason, or lead to other Adobe creativity with the data. TIFFs are apparently better understood. The (16 to 8-bit/ch) dithering that Qimage uses may lead to smoother gradients anyway, because it virtually adds half a bit to the 8-bit pipeline, without affecting resolution in a negative way.

My own printer seems to show a slight magenta offset in the head alignment under high magnification (Like some CA issue in a camera.).  I used the Auto setting in alignment, but maybe doing it Manually might be better (Although i get into self-doubts in Manual examination at times:  Is the other number better than this one?  Potshot.).

Interesting endeavor though and helps one learn how sharp - or not - their printer is.

Yes, focusing on one aspect -resolution- has its benefits, and it helps to have an objective quantification as well. Aiming for a moving target makes the process less predictable, and these cycles/mm are fixed goals. Because it is greyscale, color deviations are easier to spot. The gradient at the bottom, in the smoothest part at the bottom anyway, will also show how good the profile is at rendering greyscales with color ink.

Cheers,
Bart

[Some Guy]

Yeah, I need to revisit the Manual head alignment at some point vs. the Epson Auto mode.

Not knowing how Epson does their Auto alignment, nor knowing what inks and placement of the ink on the head's ports is another matter.  In manual, the targets seem small to what the head is capable of and maybe something there too as the head is about an inch tall vs. the tiny target of maybe 1/4" sq.  Don't know.  Their manual is sort of sketchy on manual too.

I'm not that sure of the paper thickness that Epson uses either transfers that well to the software/driver.  I can use a digital micrometer and set the thickness, but if I set a thick paper I might get a head strike or hear the "clop-clop" drag of it on the paper even if I set the paper thickness in the driver.  Perhaps any paper swells or edge curl and then so goes any alignment.  Words like Narrow, Normal, Wide, Wider, Widest and Auto are often guesses too.  I know I need to set a lot wider number based on what the mike tells me too, and not all that linear either (Else "Clop-Clop" sounds.)

Probably whatever I now use is "Good enough for an Epson."  Might not print on the head of a pin worth a damn though. 

SG

[henrikolsen]

Together with your target, I've put in the resolution test image from the bottom of this page, http://www.clarkvision.com/articles/printer-ppi/.

Printing on a Canon Pro-1000 at 600 dpi on a high-glossy paper, I notice I cannot separate 1 pixel bars (the tiniest) from clarkvisions test image, on vertical lines that is (running up/down, perpendicular to print head movement axis) - on horisontal lines they are easy to distinguish. Does that indicate the need for a head alignment (and not feed adjustment)?

Tried unidirectional printing, and that helps precision / edge separation of vertical lines.

Printer is almost new, and had a head alignment on the paper included for the task at initial setup, and has printet very little amounts yet. Perhaps head alignment needs to be paper / media type specific (guess it'll only have on alignment saved though, even it could make sense per media type / head gap).

Also notice that black contains dithering from other colours, making it even more critical with alignments. Would prefer a setting that used only black ink, but don't know of such.

Looking through a good quality 10/20x microscope.

[Bart_van_der_Wolf]

Together with your target, I've put in the resolution test image from the bottom of this page, http://www.clarkvision.com/articles/printer-ppi/.

Printing on a Canon Pro-1000 at 600 dpi on a high-glossy paper, I notice I cannot separate 1 pixel bars (the tiniest) from clarkvisions test image, on vertical lines that is (running up/down, perpendicular to print head movement axis) - on horisontal lines they are easy to distinguish. Does that indicate the need for a head alignment (and not feed adjustment)?

Tried unidirectional printing, and that helps precision / edge separation of vertical lines.

Printer is almost new, and had a head alignment on the paper included for the task at initial setup, and has printet very little amounts yet. Perhaps head alignment needs to be paper / media type specific (guess it'll only have on alignment saved though, even it could make sense per media type / head gap).

Also notice that black contains dithering from other colours, making it even more critical with alignments. Would prefer a setting that used only black ink, but don't know of such.

Looking through a good quality 10/20x microscope.

Hi Henrik,

Yes, vertical lines can benefit from head alignment, because the most prominent variable is the horizontal head movement and thus horizontal spacing. However, the alignment procedure might also involve the paper feed direction, and there may be interdependencies (due to the weaving on multiple passes). In principle, besides uni-directional printing, the head speed also has an effect since droplets will strike the media surfaces literally on-the-fly and could deform them a bit.

A second check, besides the square patches with fixed resolution in 4 directions, for asymmetric resolution in my target is the central star section. If resolution is symmetrical, then there will be a more or less disc-shaped region towards the center of the star that is unresolved (ideally up to the 600PPI resolution indicator). Otherwise, it may become a more elliptical blur area.

The nextl variable is the print medium itself, but it should blur (through ink diffusion) more or less equally in all directions (besides the earlier mentioned head travel speed).

Unclear is if/how the height of the heads above the print surface is managed with media of different weight/thickness. To play safe I usually recommend to use the same media (thickness) for alignment as the media that are going to be actually used. In general (semi) glossy media should be best for alignment purposes, assuming that they show the least amount of ink-diffusion and thus show the alignment patterns more clearly.

I suggest to look closely at how the resolution is limited in the central star region (how much more than the 300 PPI circle is resolved, how close to 600 PPI can you get?). That might give a clue as to what might cause resolution asymmetry. Fixed resolution patterns may also be affected by profile non-linearity, and may cause issues that are not noticeable in regular images (which is what the radial sinusoidal star patterns try to simulate).

Cheers,
Bart

[Ernst Dinkla]

Yes, an old thread.

If Bart is around I might be lucky. Some questions I have.

Anyone tried the printer target on the Overdrive setting of Qimage Ultimate? I know it extrapolates to 1200 PPI or 1440 PPI and the Printer settings ask not more than 600 or 720 PPI (Finest Detail) but it could improve the print according Mike of Qimage Ultimate. On a Z3200 I printed the 600 PPI target but also a shrunk 1200 PPI 65x65 mm version (so not resampled) next to it in the same print run. In the Qimage Ultimate route the 600 PPI will be upsampled to 1200 PPI, the 65x65mm target not.

From older tests with the Qimage Ultimate target I recall that when a 2x higher resolution than by the driver requested 300 PPI or 600 PPI input is used, so a 600 PPI and a 1200 PPI (shrunk again, not resampled) a downsampling to the requested PPI number happened, strong aliasing showed that had nothing to do with the 2400x1200 droplet per inch squirted resolution. Driver downsampling the cause. This was before the Qimage Ultimate Overdrive setting in the interpolation choices was available.

Now I notice more aliasing effects in the shrunk target print 4 cycles/mm patch compared to the 8 cycles/mm in the original target and on more spots. But on the other hand the "Siemens star" shows better contrast up to the RED cirkel (was 300 PPI threshold, is 600 PPI threshold in the 65x65mm version) than up to Green cirkel in the 130x130 target. On the lines not giving aliasing from the outside up to the RED cirkel, there is aliasing at several angles though, the same with the circles in the center. Counting the lines per mm shows that the lines I mention are not aliased versions half the number. It is unclear to me whether the aliasing this time is also caused by downsampling in the driver and less pronounced due to the differences between the QU print target and Bart's design. Another option is that Mike found a wormhole to get 1200 PPI to the dithering stage of the driver, bypassing downsampling. BTW in the first driver of the HP Z3100 there was a printing quality setting that asked for 1200 PPI, disappeared in later versions.

Way longer back I used to send small texts for cards at 720 PPI to the 15-17 year old Epson 9000 that did not request more than 360 PPI and the text printed sharper than by sending 360 PPI. I have to say that was text created with Acorn Risc Os applications that used sub-pixel anti-aliasing for any output, video, print, etc. Which may have coped well with a downsampling routine in the 9000 driver or the driver did not downsample.

Targets, text, are not 1:1 to photo image structures so not all is said yet about the information transfer from image to print in printing. Barts's target is probably the best for photo images we can use right now. Which brings me to another question, what would be a good set up in targets, scanning or camera copying, to measure the transfer quality of a specific printer paper system? In that system changes of papers and the measured results then. Are there elements in Bart's target usable?  Another step includes the eye acuity threshold in that transfer system but this time not a single threshold but let us say what a perfect young eye at 2000 lumen as close by as possible can resolve and the usual 300 PPI at a foot viewing distance, the RED circle in Bart's target.

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
March 2017 update, 750+ inkjet media white spectral plots

[Bart_van_der_Wolf]

Hallo Ernst,

I'll give a reaction after at least getting some sleep tonight. I'm a bit busy but do want to get to the bottom of Qimage's Overdrive setting. Earlier attempts didn't get me better resolution than 600PPI on a Canon printer, maybe I need to try another printer and/or other media as well. In the meantime, Mike Chaney also improved the printing behavior of fine testchart detail, but he (understandably) didn't want to disclose what to look for specifically when I asked him (I suspect something to do with dithering, but I'm not 100% sure).

Cheers,
Bart

[Ernst Dinkla]

Hallo Bart, er is geen haast,


I will make more prints with different settings after aligning the heads again on a sheet instead of a roll. Might make a difference.
Could send a set to you, well described, so there is more reference. Have to do work too till the end of the week at least.
Will check the Qimage printing targets as well then.


Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
March 2017 update, 750+ inkjet media white spectral plots