Atom Feed
Comments Atom Feed

Similar Articles

2009-06-09 07:43
Look Sharp

Recent Articles

2019-07-28 16:35
git http with Nginx via Flask wsgi application (git4nginx)
2018-05-15 16:48
Raspberry Pi Camera, IR Lights and more
2017-04-23 14:21
Raspberry Pi SD Card Test
2017-04-07 10:54
DNS Firewall (blackhole malicious, like Pi-hole) with bind9
2017-03-28 13:07
Kubernetes to learn Part 4

Glen Pitt-Pladdy :: Blog

R1800 Gutenprint tricks

Epson introduced the R1800 (A3 version of the R800) around 3 years ago. The main reason for me buying it (I had a 1290 at the time) was the print life and the better colour neutral B&W prints by not having the light Cyan and Magenta which then needed to be matched with the corresponding dark Cyan and Magenta (even using ICC profiles this was often a struggle). The R1800 solved this, and with Gutenprint support was a winner for me.

Paper Choices

I conducted trials of loads of different types of papers and had profiles made for many of them. My findings where that many third party papers (even the major brands) simply where not up to the job.

This was particularly the case with the matte type papers (chalky rather than plasticy coatings) where they simply did not manage to hold the same ink density if it was put down to fast. Printing at the highest resolutions (more passes for the same amount of ink) was often needed to stop ink bleeding on the paper. The only matte paper that I was satisfied with without using high resolutions for dense prints was Epson's own.

The glossy / pearl (plasticy coatings) papers all handled the ink density well, but there was a distinct difference in colour. Many of the highly reviewed third party papers turned out to be less neutral and had a blue-white (often perceived as a whiter-than-white) colour. This makes me think of tricks used by some TV manufacturers where they found that by boosting the blue slightly in their sets, they stood out in the showrooms and sold more. Could this be the same trick?

Epson photo papers are very expensive, especially for the highest weights, and while the print quality is excellent, I think they have competition for these types of paper. My eventual choice of photo paper is very reasonably priced and available in both plain Glossy and Pearl (rather like what others call Lustre): Ilford Galarie Smooth.

Both the Glossy and Pearl give very similar colour rendering, so one profile is likely to be ok for both unless you are fussy about perfect colour.

Perfect Smoothness

One thing that I have encountered with this printer with Gutenprint was strange effects with colour in blue regions. After much experimenting and examining the prints under loupes and microscopes, I came to the conclusion that Gutenprint is probably using the auxiliary colour inks (Red, Blue) differently to how Epson intended and this is the cause of the problems. This manifests itself as colour artefacts - in this example, blue and red lights on skin:

R1800 Gutenprint blue colour artefacts

On a colour test strip we see uneven colours:

R1800 with Gutenprint - colour strip (CMYKRB)

As can be clearly seen on this test strip, the blue area is uneven and the density dips briefly as we move towards magenta. This is likely what is causing the artefacts seen above.

The auxiliary colours are much darker and have far more visible dot patterns:

R1800 CMYKRB vs CMYK dot visibility

I suspect that Epson intended this printer to operate in CMYK mode in the more neutral area, and only use auxiliary colours at higher densities to extend the gamut. Gutenprint seems to have chosen to use them as primary colours (ie. even at low densities) which results in very visible dot patterns in lighter blue and to some extent red areas.

Combining this with the uneven rendering of the colour transitions when using the auxiliary colours, I have decided to disable the auxiliary colours and use the printer in plan old CMYK mode. I did have a quick look at the source code to see if I could change the rendering of auxiliary colours to only be used at high densities, but this looks like a lot of work which I don't have time for. Switching to CMYK yields less visible dots and smooth colour transitions (ICC profiles need to work less hard) with the loss of extremes of gamut (I can live with that):

R1800 with Gutenprint - colour strip (CMYK)


Having spent years messing around trying to get good colour in both prints and monitors, I have given up and done what everyone else who wants good colour does: colour management.

Colour management is highly effective at getting consistent colour across different devices (cameras, scanners, monitors, printers, projectors etc.), but does require care and a degree of understanding to use it correctly. This is often where things go wrong - people don't take the time to learn how to set it up right, they don't verify their setup against a known good colour reference, and they don't maintain the setup. The results is that many people have had bad experience with colour management and dismiss it outright.

As I colour manage everything, the printer setup is based around using colour management.

Basic Gutenprint Settings:

Printer model: Epson Stylus Photo R1800
Print Quality: Best
Media Type: Premium Semigloss Photo Paper (Ilford recommended)
Media Source: Standard
Ink Set: Photo Black
Enhanced Gloss: Your choice. I use it for anything that goes on display.
Borderless: ON
Resolution: 5760 x 2880 DPI
Printing Direction: Automatic
Interleave Method: Standard
Ink Type: Four Colour Standard (see above)
Adjust dot size as necessary: OFF

Adjust Output / Print Colour Adjust settings:

Image Type: Photograph
Brightness: 1.000
Contrast: 1.000
Saturation: 1.000
Colour Correction: Uncorrected (that's the profile's job)
Dither Algorithm: Adaptive Hybrid

Everything else on it's defaults except:

Composite Gamma: 0.900
GCR Transition: 1.000
GCR Lower Bound: 0.300
GCR Upper Bound: 0.980

Colour Profile

Various versions of Gutenprint are subtly different in their colour rendering, so if you are not using exactly the same version then you may find that your colour is a bit out. Also, printers vary from one to another, and I can sometimes also pick out slight differences when one cartridge is nearly empty and the others are all full.

Also, keep in mind that with the widespread use of CFL lighting these days, you simply don't get a smooth illumination spectrum (CFLs often give a very peaky spectrum) with indoor lighting and this interacts with inks resulting in metameric failure and different colour being seen under different lighting conditions as a result.

Download: Gutenprint 5.2.0rc1 Epson R1800 CMYK with Ilford Smooth Glossy paper profile

Generally I print from Gimp and simply converting the image to this profile (Perceptual without Blackpoint Compensation or Relative with Blackpoint Compensation depending on what you are trying to achieve) and print the image with the above settings. You may want to duplicate the image before doing this so that the original is preserved in the working colour space.

This profile was created using Argyll which in my experience walks all over expensive commercial systems when it comes to giving results that look most true to the original. This has particularly been the case with printing at extremes of gamut where the commercially made profiles shifted colours, but Argyll managed to render the colours much more true to the original.

Comparison between profiles from X-Rite and Argyll

These where produced simply by converting the original image to the profile (Perceptual) and then back to sRGB. As is clear (especially in the top left), is that the out of gamut blue/magenta regions simply get completely converted to blue by X-Rite, but Argyll keeps the colour true, but just limits the saturation to the printer gamut.


Update 15/9/2009

It's taken me a while to get round to putting this up:  This entry has had a fair amount of interest and I was both surprised and honoured to get a personal email from Robert Krawitz, the project lead for Gutenprint.  I found what he had to say regarding these printers insightful and may be of interest to others hitting this blog so here it is:

Subject: Gutenprint
From: Robert Krawitz <email removed>
To: Glen Pitt-Pladdy <email removed>
Date: Sun, 9 Aug 2009 22:57:29 +0100 (BST)


I am the project lead for Gutenprint. I read (what I presume to be, since there's no email address) your blog entry "R1800 Gutenprint tricks" with no small amount of interest. I have some comments on this and perhaps we may want to discuss this further.

First, a brief explanation of the DeviceN ink generation and color conversion performed by Gutenprint. Gutenprint does the color conversion in HSL space (or really HL+K -- hue, luminosity, and black level). This normally produces smoother gradients than typical RGB or CMYK profiles.

The DeviceN ink generation is done in a variety of ways. For typical light inks (CcMmYK, for example), we do CMYK separation with parameterized GCR and then split the channels based on the channel level and amount of ink. For the odd printer like the R800/R1800 and R1900, the method is different -- we do the conversion based on HL (i. e. we only look at the non-gray portion of the color). Unfortunately, this results in some odd effects in the region where the colors kick in. I've observed the same kind of thing you have but have not been able to resolve the issue.

I suspect that you're right that the intention is that the red and "blue" (which is really purple) inks are intended to be used only in dark, highly saturated regions. Unfortunately, I have not been able to devise an algorithm that does this without having other undesirable effects. The CMYKRO in the R1900 is even stranger -- why is there an orange ink between the already non-primary red and the primary yellow? My assumption is that it is intended to help with skin tones.

Your approach of using 4-color standard (CMYK) may be best for most people. I've observed that the primaries in this printer produce very weak reds, though, but perhaps profiling it would yield better results. I suspect the only way to use the RB inks would be with a carefully constructed DeviceN profile.

Personally, I'm not a big fan of any of these printers. Without the light inks, they need to use very small drops to achieve smoothness, which means printing at very high resolution. I much prefer printers like the R2400/R2880 or R1400, which use conventional CcMmYKkk or CcMmYK inks. Despite the much larger minimum drop size
on the R2400 and R2880, they actually produce smoother output, even at 1440x720 DPI.

I'd like to improve the output on the R800 family for the next major release of Gutenprint (we're trying to hold colors stable in the 5.2 series, and have better testing procedures to ensure that). But I'm reluctant to make stabs in the dark that might simply result in worse problems.

If you want to include this in a blog entry, please feel free to.

I have also acquired a automatic spectrophotometer and am hoping to start offering custom printer profiles. Initial tests have given superb results (far better than my profile here). I am also hoping to be able to offer a "Open Source" profiling service (probably under Creative Commons) in coming months, but more about that another time.