thebestcpu
Fan of Printing
- Joined
- Dec 8, 2024
- Messages
- 102
- Reaction score
- 60
- Points
- 60
- Printer Model
- Epson SC P900
I am not an expert in color measurement or ICC profile creation. I’ve recently revisited this field—working with an i1Pro and ArgyllCMS—as part of some experiments. My day-to-day printing relies on stock ICC profiles, so this is a learning journey.
That said, I encountered something during my testing that may be worth raising for consideration.
In my experiments with the Epson P900 (specifically with Black Enhanced Overcoat enabled), I observed measurable spatial variation in output. Grayscale values printed at the top of the page were consistently 1 to 1.5 ΔL darker than those printed approximately 8 inches lower. In expanded color tests, I measured ΔE2000 differences between 2 and 3 for certain patches depending on their vertical position on the page.
While such variation may not be noticeable in typical photo printing—where changes are gradual across large areas—it occurred to me that this could have implications for the design of profiling targets, especially those with finely stepped luminance ramps.
For example, using a high-resolution target with 128 steps from dark to light, the ΔE between adjacent patches may be less than 1. If spatial non-uniformity introduces noise greater than that, then randomizing the patch layout (a common practice to average out localized printer errors) might inadvertently introduce variations that exceed the intended step resolution—potentially disrupting the smoothness of LUT construction in ICC profiles.
It's entirely possible that profile generation tools already compensate for such artifacts through smoothing or statistical methods. My intent here isn’t to challenge established practices but to share a data point and raise a question I haven’t seen addressed: How is spatial printer variation—distinct from measurement noise—accounted for in target design or profiling workflows?
I’m not pursuing this further, but I’ve included a simple test and measurements from my Epson P900 to illustrate what I observed.
Test Description
To estimate top-to-bottom spatial variance, I created a test image with vertical stripes: six grayscale levels followed by three stripes each for the primary and secondary colors, with spaced luminosity.
The image was printed using Photoshop and scanned on an Epson V800 (also scanned in reverse orientation as a bias check). From the scan, I isolated 24 patch regions from the top and 24 from the bottom, averaged each, and calculated the ΔE2000 values between corresponding pairs. The first row of values under the image shows ΔE2000 results. The second row shows the difference in L values (bottom-top) to show directional shifts.
Thank you for taking the time to review this. I welcome any thoughts or clarification and appreciate your insights into whether this phenomenon is already accounted for—or if it's worth exploring further.
John Wheeler
That said, I encountered something during my testing that may be worth raising for consideration.
In my experiments with the Epson P900 (specifically with Black Enhanced Overcoat enabled), I observed measurable spatial variation in output. Grayscale values printed at the top of the page were consistently 1 to 1.5 ΔL darker than those printed approximately 8 inches lower. In expanded color tests, I measured ΔE2000 differences between 2 and 3 for certain patches depending on their vertical position on the page.
While such variation may not be noticeable in typical photo printing—where changes are gradual across large areas—it occurred to me that this could have implications for the design of profiling targets, especially those with finely stepped luminance ramps.
For example, using a high-resolution target with 128 steps from dark to light, the ΔE between adjacent patches may be less than 1. If spatial non-uniformity introduces noise greater than that, then randomizing the patch layout (a common practice to average out localized printer errors) might inadvertently introduce variations that exceed the intended step resolution—potentially disrupting the smoothness of LUT construction in ICC profiles.
It's entirely possible that profile generation tools already compensate for such artifacts through smoothing or statistical methods. My intent here isn’t to challenge established practices but to share a data point and raise a question I haven’t seen addressed: How is spatial printer variation—distinct from measurement noise—accounted for in target design or profiling workflows?
I’m not pursuing this further, but I’ve included a simple test and measurements from my Epson P900 to illustrate what I observed.
Test Description
To estimate top-to-bottom spatial variance, I created a test image with vertical stripes: six grayscale levels followed by three stripes each for the primary and secondary colors, with spaced luminosity.
The image was printed using Photoshop and scanned on an Epson V800 (also scanned in reverse orientation as a bias check). From the scan, I isolated 24 patch regions from the top and 24 from the bottom, averaged each, and calculated the ΔE2000 values between corresponding pairs. The first row of values under the image shows ΔE2000 results. The second row shows the difference in L values (bottom-top) to show directional shifts.
Thank you for taking the time to review this. I welcome any thoughts or clarification and appreciate your insights into whether this phenomenon is already accounted for—or if it's worth exploring further.
John Wheeler
