RSS Feeds
Ultra-Precise Reproduction Accuracy
Mike McNamee tunes Lukman Sinclair’s new Epson 7880 to colour perfection
The brief we set ourselves was relatively simple to set out – colour profile a new 7880 for art reproduction specialist, Lukman Sinclair, squeeze the thing until it squeaked and then make it deliver breathtaking colour accuracy! This is not for the faint-hearted and ordinary users should not be tricked into thinking that they need this level of fine tuning, printers such as the Epson 7880 are good for 99.9% of commercial jobs right out of the box.
Botanical Reproduction
No we don’t mean the birds and the bees, we have written about Lukman’s specialisation before. It is the reproduction of very high-quality, colouraccurate copies of world-class artwork, made originally for organisations such as the Royal Horticultural Society and Kew Gardens. The demands are high. The output usually consists of a single ‘perfect’ reproduction, so that the original may be lodged in one of the learned societies’ collections. The difficulty arises because the artist is able to make a direct comparison between the original and the reproduction, a task many artists can perform at levels of discrimination that match industry-standard colour measuring devices. The artists themselves may not express the colour difference in ‘colour management speak’ but they sure know when the colours match! With a formal training in photography and a lifetime of mixing his own artist’s palette, Lukman is well-placed to perform the assessment of the reproduction and makes careful, selective colour adjustments using layer masks. Our job was to tune the profile so that the colours he created from the scan were a perfect match to the print. Along the way it gave us the opportunity to critically appraise the new Vivid Magenta ink set that comes with the 4880 to 10880 series of Epson printers.
That's the colour I want!' Epson's Neil Mackenzie (right) and Lukman Sinclair share a joke while McNamee sets up the profiling kit.
Epson Ultra Smooth Fine Art Velvet
To set a benchmark before we commenced profiling and to give us confidence in the printer set up, we checked out the 7880 using the Epson profile which comes with the drivers on the CD. Epson have spent a lot of time and trouble on their latest generation of profiles and this one dropped in at 5.7DE Lab and 2.62 ΔE2000. This sent us scurrying to the database. We now have data on more than 1,500 printer/paper combinations and of these we have more than 350 from the Epson 3800. This has enabled us to build a picture of how Epson ink sets perform across a wide range of media and printers. The average is 3.4ΔE2000 for all data on the Epson 3800 and this includes both the matt black and photo black variants. A value of 2.6 therefore is quite a significant result which was not bettered by any of the recent trials on art papers. Things thus started on a bit of a high note!
Examining the data in detail showed that the bulk of the colour errors were in saturation component and that the bulk of the greys' error lay in the lightness component. This is typical for a matt paper as the Dmax is usually around 1.5 to 1.6 and the gamut is constrained by an inability to achieve very rich, dark colours – too much black ink has to be added to get the tone density, leaving the colour a little degraded in saturation. The truth is though, that in these very dark tones, the colour is difficult to detect anyway. The only other point we noted was that the in-built profile created a rather sharp transition from the powder blues into the blacks.
However, this extremely bright paper created a sparkling quality print that would be regarded by the vast majority of users as flawless. Another key feature was the subtle differentiation of the whites right out to between the 252 and 254 RGB points; we don’t ordinarily get that differentiation.
Round one then put us in a very positive frame of mind as we progressed to the paper that Lukman’s clients prefer the most – Hahnemühle Photo Rag.
Hahnemühle Photo Rag
Three profiles were prepared, one with the X-Rite Pulse (on a 343 patch target), one with 729 patches from Monaco Profiler (on a DTP 41) and finally a tuned profile (see call-out box for tuning methodology). The modification required for the profile was a simple, all-round tweak to the saturation, to boost it by 4% in the middle of the range.
The gamut volume measured at 544,054 which may be compared with that from the Epson 3800 on the same paper of 541,293; this, in fact, is a very small difference, possibly brought about by the expansion of the gamut with the Vivid Magenta ink, but almost certainly compromised by the matt paper – the difference would show up more on a full gloss or lustre paper.
The tuned profile returned an error of 4.5LabDE/2.24 ΔE2000 which is the best result ever from all of our matt black ink results. The average of all results using a 3800 and matt black ink (ie on a matt paper by definition) is 3.4 (actually the same as the overall average) with a range from 2.4 to 4.3 ΔE2000. This is all for bespoke profiles. The main component of the error is usually the lightness channel and this was so in this case also, where it was actually 50% larger than both the saturation and hue errors. This reflects the limited Dmax of the matt papers.
Detailed Analysis – the foliage greens
For the type of work that Lukman specialises in, the accuracy of the foliage colours are important and almost universally present (flowers vary in colour but most leaves are actually quite similar in colour values). To investigate the detailed performance we made a special set of test swatches, selected to encompass all the colours we measured for the artists’ medias plus a little leeway, further into the gamut. The clusters of colours were arranged in four brightness levels of 75%, 60%, 45% and 30% Lab lightness values.
The simplified data table shows that the error progressively rises as the Dmax of the black is approached, a value of 30% is only 10% shy of the maximum black that the ink/paper combination is capable of. Looking in more detail at the plot of errors across all the leaf tones also shows that the error increases with saturation that is towards the top left of each set of swatches. Again this is to be expected, especially as the residual errors we left in the profile was the saturation component, a further profile refinement might have improved things but just pinpointing the error distribution took half a day’s work and we had returned home by then! The errors in the swatch sets at 75% and 60% brightness are just on the limit of detection for a human eye in controlled conditions. To put the error into perspective, a change of light source from D65 to Tungsten A was creating a difference of 17 ΔELab, that is about eight times higher than the error we achieved from the print. Even shifting from D65 to D50 was showing a difference of 5 ΔELab.
The bottom line to the analysis is that the colour error from the Epson 7880 was by now so small that it would not be practical to refine it much further, the limitations of metamerism were now controlling the result.
CAPTIONS: At the top, the target of green tones has been superimposed with the error values measured from the adjusted profile for Hahnemühle Photo Rag (HPR). These indicate that there is an overall fall-off in accuracy of reproduction as the tones become darker and that the less saturated colours are more accurate. On the right the gamut for the HPR is plotted in 3-D with the values of the dark greens seen outside the gamut. The pair of graphs on the right show very close agreement in hue and saturation (ie colour) at 75% brightness, where the colours are some way inside the gamut boundary. In contrast the lower graph shows the colour values being desaturated as they are pulled down into the gamut (the red line shows the movement of the most saturated green in the set. The histogram at the top of the next page confirms this trend across the entire Macbeth set of swatch colours. The yellow, purple and blue bars show the hue, saturation and brightness errors, with the data for the 7880 being on the right and the average data for all 3800 trials on the left. Note that the errors in the lightness component are not improved because the limiting factor is the density that the matt paper can support. These data do not imply that there is a large difference between the 3800 and the 7880 but suggest that the Vivid Magenta is doing its work.
BELOW: The pastel study by Deborah Lomas offered a challenge for the new Vivid Magenta ink set of the 7880, although the out-of-gamut plot from Monaco Gamut Works indicates that the orange tones are the issue! By hand-plotting the most saturated colours from the image we found that they were just on the gamut boundary, not outside it – these utilities are not as precise as we would like.
Boy in Turban – squeezing the magentas
The artwork for this sequence of tests was an original oil pastel by artist, Deborah Lomas. It was an ideal image as it contains vivid magentas, just the tones that the new ink set was designed to reproduce accurately! The original was first scanned on the Epson GT 10000 with a bespoke profile. Lukman then refined the scan with masked adjustment layers to correct the metamerism from the scanner and reproduce the pastel colour on screen as closely as possible. We then examined the image using a variety of colour-management tools to see just where we stood in relation to the 7880 gamut as defined by the profile for HPR. Monaco Gamut Works indicated that all of the magenta turban was in gamut but some of the orange back ground was out of gamut. Photoshop indicated that the entire image was in gamut. Plotting the colour values by hand showed that the turban was perilously close to the gamut boundary but just inside and that the same was true of the orange background. We did not have a bespoke test set of swatches but the closest visual match from the FOGRAGRAcol contract proof set had an error of just 0.99ΔE2000.
Analysis of the data simply legitimised what our eyes were telling us, the difficult magentas of the turban were reproduced so as to be quite indistinguishable from the original – it was an impressive demonstration
Overall
At the end of the trials we were pleased with the results we had acheived and satisified that the 7880 was tuned as close as we could reasonably expect to get it. The residual errors are now almost exclusively due to variations in scanner metamerism interacting with the artist's pigments, something that the person reproducing the image can do little about. As for Lukman, he is delighted with the new machine as well he should be, close as it is to colour perfection!
Preparatory Work
We were reminded of some problems that we had to solve back in October 2004 with the Windows Spooler. This is worthy of a mention as it occurs when you step up to larger prints, possibly the reason you have gone for a 7880 or 9880 in the first place. The symptoms are the ejection of a print just before it is completed, leaving the uses with a very expensive partial test print – it usually involves A2 size prints or larger so the wastage-aggravation is high! The recommended solution is to do the following:
1. Uncheck the ‘Enable Advanced Printing Features’ in the dialogue box found by clicking Printer>Properties>Advanced.
2. If the problem has been persistent for some time before you read this, go to your TEMP folder and empty it, it is likely to hold the remnants of your failed printing and can itself cause problems.
This procedure is claimed to solve 99% of spooler issues on the larger printers. Our own experience shows that it has been an issue with the 24” machines but we have never had a failure of the spooler using the Epson 3800 or 4800 even though the Enable Advanced Printing Features has always been checked.
To judge from the comments on the web forums, Microsoft have learned very little in the intervening period between October 2004 and today. Vista forums are also liberally scattered with bitching about spooler issues.
For further information start at http://www.steves-digicams.com/techcorner/ february_2006.html or the original article in Professional Imagemaker may be found at http://www.swpp.co.uk/professional_ imagemaker/print/index.htm
The Tools
We used Monaco Profiler Platinum to fine tune the profiles. This is a specialised procedure for the software itself provides very specialised tools! You have to be able to audit and measure the output, sit and analyse the data, and then make educated guesses as to how to tweak the various colours and curves – it is not for the faint-hearted and takes the best part of an afternoon for each round of corrections. It is an iterative process sometimes taking three rounds, although you sometimes get lucky and do it in one round!
The program enables the user to set up their basic profile and then edit it either globally for Lightness, saturation and RGB curves or using the more selective colour options. With selective colour tuning you may pick a target colour either numerically or with a colour dropper and then choose how to shift that colour within the gamut. It is possible to surround the chosen colour with a protective bubble so that other parts of the gamut remain unaffected. The shape of the bubble may be varied in both the lightness and colour co-ordinates.
The entire process requires tight control, consistent methodology and a clear head; this is not a job you should do at the end of a long day or after a socially strenuous weekend!
Photo Quote: The contemporary artist...is not bound to a fully conceived, previsioned end. His mind is kept alert to in-process discovery and a working rapport is established between the artist and his creation. While it may be true, as Nathan Lyons stated, 'The eye and the camera see more than the mind knows,' is it not also conceivable that the mind knows more than the eye and the camera can see? - Jerry Uelsmann
