Resampling for printing – summary

Have you been looking at the last couple of weeks’ posts and been wondering what to make of it all? So have I. Here’s what I think:

The marking engine (printer engineering talk for the collection of parts that lays down the ink, or toner, or whatever) of the Epson 3880 is capable of consistent dot sizes smaller than one-thousandth of an inch on glossy paper. I see no reason to doubt the manufacturer’s assertion that the printer is capable of placing those dots on a 2880×1440 dot/inch grid. With conventional error diffusion techniques, the printer should be capable of marginally resolving 1440 line pairs per inch in one direction and 720 line pairs per inch in the other. Used with the Epson driver, it cannot. Downhill with a tailwind, it can resolve 360 line pairs per inch, and it doesn’t do that well. The best dependable resolution is 180 line pairs per inch. How come? Two reasons:

First, the driver resamples the image before it or the printer applies the error diffusion. When the Finest Detail option is checked, the image is resampled to 720 pixels per inch, and otherwise it’s resampled to 360 ppi. The resampling algorithm is nearest neighbor.

Second, there is enough noise (dither) added to the error diffusion algorithm that a single pixel line at 720 ppi occupies about twice that width on the paper.

Why did Epson do things this way? I can only speculate. I think that they have biased their halftoning algorithms to favor smoothness over resolution, and I think they want to avoid things that look artificial at all costs. Also, the resolution that I’m talking about is only obtainable for the smallest of the three ink drop sizes; the way to make sure you get only the smallest drops is to use the 2880/1440 ppi resolution setting.

The nearest neighbor resampling performed by the driver can produce some really ugly effects. You want to resample the image yourself to the resolution that the driver uses for that dot pitch before sending it to the driver. Lightroom makes this convenient, but the results are softer than they have to be. The Photoshop bicubic interpolations produce reasonably good results.

There is a program called QImage that can resample with many algorithms, including two proprietary ones: Hybrid and Hybrid SE, which produce a combination of smoothness and crispness that’s better than what you can get by using the Photoshop interpolation algorithms.

Checking the Finest Detail box produces better results only in photographically unusual situations.

Convert to the printer’s RGB color space in Photoshop, Lightroom, or whatever image app you choose. Do not let the Epson driver perform the conversion. Turn color management off in the driver.

Using a RIP instead of the Epson driver might produce better results. I may try that if there’s enough interest. Using a third-party image scaling plugin within Photoshop may produce as good or better results as QImage. I don’t plan to try that since it wouldn’t fit my workflow.

I would expect that most of the above applies to other Epson printers, but I don’t know for sure. I have a 9800, but I use matte black ink and matte paper in it, and don’t want to switch inks twice to find out how the printer performs in a mode in which I’ll never use it. I could do some tests with matte paper, but they wouldn’t be comparable to the glossy tests I did with the 3880.

I have questions about the Epson algorithms. Color management? Conversion from RGB to CcMmYKkk? Resampling? Error diffusion? I would love to know what they really are and what gets done in the driver vs the printer. Anybody?