Bart, if the target was saved as a tiff, the ppi setting of either 600 or 720 remains sticky.
Indeed Samuel, I also found that out this morning.
Looks like you have discovered something weird about the way PNGs are tagged with the image resolution value. What a funny problem, but also a very serious one. The 600ppi PNG target still opens as 599.999 ppi instead of 600ppi in Photoshop CS6 for me. I'm not sure how kers and you got it to show up as 600ppi. I'm going to resave them as tiffs for now...
I've uploaded two new (tweaked) versions, as TIFFs, and on my computer they open in Photoshop CS6 with the intended/correct PPI indication. The links in the original post have been adjusted to point to the new TIFF versions. While there is nothing wrong with losslessly compressed PNGs per se, apparently Photoshop uses some strange(?) logic that messes up the PPI tags in PNGs but not in TIFFs, at least at these specific sizes.
Here are the links to the updated/tweaked TIFF versions, I'll remove the PNGs to avoid confusion:
Test target for 600 PPI printers (e.g. Canon / HP)
Test target for 720 PPI printers (e.g. Epson)
I've used ZIP compressed TIFFs that may not be readable in software that uses non-standard or old TIFF libraries. If that poses any problems I will change that to much larger uncompressed versions, and put them in ZIPped archives.
The current tweaks are mostly about an even higher quality of the sinusoidal patterns with the 4 quadrants of uniform period. I also attempted to reduce the difference in average density of the horizontal/vertical and diagonal patches of the highest possible resolution, at the top left corner. The higher quality should reduce the chance of aliasing when everything else is perfect even further. Any printed misalignment should therefore still stick out as a sore thumb, and only be caused by that misalignment, or by resampling and sharpening.
The other patterns in the target will be reviewed and possibly adjusted/tweaked at a later stage, but for the sinusoidal based ones that would take an enormous amount of computation time because they are larger in size.
I think you have partially won me over to the use of sinusoids in your printer target. I agree they are useful, and uniform high contrast pixel lines cannot reveal the full story. I will pay close attention when you post examples from your printers and comment about them. I am curious to know what other issues these patterns can reveal.
The sensitivity of minute misalignments/distortions that manifest themselves as significant aliasing artifacts is used in other fields as well. It can be used in accurate angle measurements with mechanical inclinometers, or for measuring dimensional expansion or contraction. It is somewhat related to the higher accuracy one can achieve with using
Vernier calipers for distance measurements, and the Vernier acuity which we can exploit by printing at 600 or 720 PPI instead of lower settings. While normal visual acuity is on average often limited to something like 300 to 360 PPI, we are able to see much higher resolution when Vernier acuity is taken into consideration.
We can use the properties of
moiré patterns to search for the cause and magnitude of the misalignments. The simplest is the head alignment and paper travel alignment which should produce similar density in the horizontal/vertical sinusoids, but there may also be issues with paper that is not fed squarely but at a slight angle due to some slippage or uneven friction.
I just tried resampling the target in Photozoom Pro, it did not give a near perfect sinusoid rendering of the patterns but rather enlarged (obviously lol) and made the low frequency aliasing patterns a great deal more apparent. It is only these patterns that are irritating me at the moment - one needs to learn what patterns are representative of the actual data in the target, (just at a different size like you said) and what patterns indicate resampling, or misalignment or additional sharpening or dithering effects etc.
Yes, PhotoZoom Pro will accentuate high contrast edge detail by
adding resolution where there was none before. When measured accurately, the resolution that is created, e.g. by upsampling to 600 or 720 PPI, exceeds that which was available in the original file. While that may show as a bit strange on the test target because that level of detail was not present in the original sinusoids, it works very well on natural images which are obviously much more random/organic (and noisy) than these mathematical constructions. Down-sampling in PhotoZoom is not very good, but for upsampling the results can hardly get better.
Can't even imagine the computational load these sinusoidal calculation require! May I say a big thank you for Above 5 cycles/mm frequency, it looks like the printer is unable to perfectly resolve the sinusoidal patterns, partly due to dot size, partly due to the density of the inks.
Well, the sinusoids are not the hardest part, but the way we point sample or area sample at the pixel level can be computation intensive. But that's what we have computers for. All we have to do is be patient while the calculations are running. I don't think printers or cameras will be able to get it 100% accuretely, unless we can use
much higher resolutions. But that's not very practical given all the other variables like paper transportation and color dithering, and bandwidth limitations when sending the print-data.
I am not sure how looking at the printed target could reveal that the printer pipeline is capable of 16 bits precision? Assuming no dither was applied in the process.
Other than comparing the output of an 8-b/ch and a 16-b/ch version of the target, I'm not sure either. I may need to add another tricky pattern to the target to amplify those differences. There may be some utility in the two (1, 2, 3 line wide) patterns at the top right and bottom right of the neutral star background, and the steps in the gradient. Of course colour dithering is also something that can benefit from 16-bit print pipelines, and those tonalities are more vaguely related to resolution.
Cheers,
Bart