What is the optimal output resolution to a printer?

I’m convinced it depends on the printer but there are a lot of issues involved independent of the printer but since I use Epson printers (most lately the R800 pigment ink printer) I will speak mainly to that issue.

In a normal continuous tonal range print the human eye can resolve something like 240 dpi (dots per inch). That figure varies greatly depending on the contrast of the edges we are trying to resolve. For many years I have done a lot of work in Illustrator and other structured drawing programs and it has been obvious that non-dithered output can be preferable and that output has to be very high to maximize the appearance to a smooth curve or diagonal line. I have to admit that in real world photography this type of output is rare if nonexistent. Output significantly higher than 300 ppi (pixels per inch) is needed to approximate a continuous non-digital edge. Higher is better but how high is enough to maximize the capabilities of the printer? It takes about 60 seconds to at least plant the seed of doubt regarding the purported Epson maximum file resolution of 360ppi so I don’t understand why part of this dialog is ongoing. It’s so easy to produce a hands-on example. Draw a couple of curved and angled structured graphic shapes (as in Adobe Illustrator or Corel Draw) and open them in PS at various resolutions. Print them at the Epson printer’s highest print quality; in the R800’s case Photo RPM. Try it. Don’t just talk about it. The printer can without ANY doubt, resolve more than 360 ppi. It is irrefutable and obviously irrefutable as there is a big difference between 360 and 720ppi in output. It’s obvious when rastering vector graphics… I’ve been doing this for almost 20 years and its as true now as it was in those ancient times. The maximum hardware (or native resolution) of an Epson printer appears to be 720.

At some unspecified time in the past, a statement I made in one of my Dpreview posts was challenged by Mike Cheney, the creator of Qimage. While a lot of issues were involved, Mike convinced me that under certain circumstances bad things happen when you let the Epson (and presumably other) printer drivers upres to the printers native resolution and that I would be better doing the job myself in PhotoShop. Mike suggested I look (my remembrance) were there are instances of high contrast photo imagery that has a repeating pattern. If the pattern is of the right frequency (i.e. you have the bad luck to have it in your photo) moiré rears it’s ugly head. Moiré is a rainbow-like striped effect that is caused by interference patterns between input and output frequencies. For my tests I chose the file; Mike had no control over my experimentation other than his original suggestion that I was barking up the wrong tree with my statements. To ensure that I had a file with the “right” frequency I chose a resolution test chart from a high quality camera on the assumption that it would have an extremely wide rage of frequencies where moiré would somewhere show up. It manifested itself in a few places at 300 ppi quite strongly in my first test print. When I interpolated my test files to 360 dpi (an even fraction of Epson native res) and printed, it minimized its effect as opposed to lower ppi. 720 ppi almost totally eliminated moiré on the same file if it was not inherently present. Real world examples of these frequencies would be clapboard siding, picket fences or hair of the appropriate dimensions. I am leery as to how often these situations manifest themselves in our photographs but I am convinced that they occasionally occur for some of us more than others. It will happen to all of us if we print enough.

I cannot insist that there is not some other magic ppi that will eliminate this moiré as well. However I find it too much of a coincidence that without his being correct, Mike’s suggestion miraculously led to the elimination of a manifestation of photographic illness in a file of MY choosing. For all these reasons I see why Qimage works well.

The problem is we are generally not printing vector graphics or anything that remotely resembles them. People will dance around this issue like it is the Ebola virus and perhaps rightfully so. The printer driver is like a magic black box that we throw information into and a picture pops out via another magic black box called the printer. What most people think goes on in this magic place is absurdly simplistic. 99+% of people, myself included, fall into this “most people” category. What is clear to me and most of us is that the inkjet printer driver is using a relatively few pixels per inch to generate a huge array of printed dots per inch on paper. These dots are of various sizes and colors and are intermingled to approximate visually “what is in the picture”. All of us are trying to provide the printer driver with good information that will allow us to achieve the “best “ image.

This is how I went about my experimentation and the results you see below. I downloaded a Canon 1Ds test chart from DP review, used levels to increase contrast to reasonable amounts, rescaled to 300ppi and used modest USM to prepare the base file. I then resampled this image to 360ppi and 720ppi. Out of PS I printed all 3 variants at PhotoRPM on high quality glossy paper on my R800. Between 300 and 360ppi there was a HUGE change that was visible to the naked eye even under casual observation. I had to reprint and reprocess these images a number of times because of subtle “illness” in the image itself. Here’s where it gets really sticky. To allow the reader to see what I printed, I had to scan the resultant prints. I did this at an appropriately high resolution and interpolated them down to approximate what I see with my eyes at about 12 inches away in decent lighting. The resolution isn’t there but as a general rule this link gives a sense of what the printouts look like. Clearly I chose a section of the test chart that exhibited nasties. Human nature being what it is I suspect that a number of people will not even bother looking at the image, as they don’t want to believe this is what happens under certain conditions or they think they know and don’t want to be bothered.

I’m assuming that by the nature of the Internet all viewers will see this image at 100% otherwise all bets are off. Moiré is omnipresent in the majority of the 300ppi portions that I’ve boxed in blue. In the 360 and 720 ppi portions boxed in red, the 720ppi image is ever so slightly better in regard to moiré. Here the moiré manifests itself as a hint of a vertical shadow band in the web image as well as in the real printout. The sections boxed in yellow are interesting and significant because that moiré is inherent in the file itself. I ran into this on a couple of resolution charts and of course it’s very subtle at the actual size. The R800 does a pretty good job of reproducing the moiré in this area, perhaps even accentuating it slightly.

No matter what the technical reason, the Epson driver has a difficult time dealing with 300dpi (and presumably other pixel densities) under certain frequencies of high contrast repeating patterns. It doesn’t take a lot of imagination to foresee bad things happening to real world photos at 300ppi output. Draw your own conclusions. This is just more fuel for the ongoing debate I suppose.