Scott Martin’s 3up evaluation image (available at http://www.on-sight.com/downloads/) has on the left side grangers that gradate from pure black to solid colors, and from solid colors to pure white.These gradations from white to solid color and then to black are super valuable and can be reproduced with perfect smoothness.
Like other grangers, much, or most really, of the colors are out of the printer's gamut. The appearance of smoothness, even at gamut edges, is a side effect of the shape of the gamut and shifts in luminance, hue, and saturation as the printed color travels along some portion of the gamut boundary. Almost all the information from these is about how profiles map out of gamut colors. Most of my printing is with colors in gamut using Rel Col or Abs Col. Profiles almost always tell you that they are printing in gamut colors very accurately whether they are or not. So one is left printing in gamut known color patches, measuring them, and running the statistics on the set. And that's what is done regularly by people to check their printer accuracy.
The problem is that it doesn't test in-gamut smoothness.
I wonder if you could render similar gradients, in AdobeRGB with this math? I feel like there is value in comparing smoothness from an AdobeRGB if we are reproducing images from AdobeRGB.
Also, if you were to evaluate the printer itself you would need to print a target without an ICC profile right? If you are printing or rendering these targets with an ICC profile you are comparing both the printer quality but also the profile quality. So if you’re seeing problems in the Green as you mention, it’s quite possible that’s a problem with the profile but not the printer for example.
The green was an observation about a subtle difference in the way my 9800 v 9500 prints. The 9800 is better overall, especially at lower luminance with a black point about half that of the 9500 and a much large gamut overall at low L values below about 35. The former is because the 9500 has a "green" ink cartridge while the 9800 has more neutral gradations which seem to yield a larger low luminance color. The differences are quite subtle. The posted images can be compared in Photoshop by flipping back and forth over them. Since the colors are all in gamut of their respective printers the differences are pretty small. Also, view proof selecting relative colorimetric will yield no change with the printer's profile and little change with others and only there when the colors exceed the other printer's gamut.
You need a profile because printer RGB interfaces are not very linear and so printing RGB grangers in device space with one printer can produce very different results than another printer. Profiles let you compare smoothness and do apples to apples comparisons.
The downside is that comparing in-gamut smooth grangers between printers limits one to the smallest gamut in both printers.
One idea you mentioned, which I think would be useful, is to do this in device space. By definition, device space is in gamut. Then one could make a gradient where each color hue was at the maximum saturation decreasing to some neutral value. Then changing the hue angle dynamically so that dE 2000 moved at a constant rate. This doesn't produce constant L values though and they would vary quite wildly with hue. The images I posted were all with constant L values which makes it easier for me to look in a region I'm suspicious about if I come across banding or gamut clipping in a print.
