Glad to here it!
I assume you do realise, donāt you, that you risk being branded as a Piezography Heretic?
This doesnāt really gel with my experiences, but letās wait and see your forthcoming workflow posts before commenting any further.
There are benefits to all the various methods of ālinearization.ā For one, Iāve been using studioprint since 2004 (primarily for itās Piezography) and it is NOT a linear slope. Nor is it drastically as āperceptualā however there are benefits to studioprint over a straight-line/gamma22 method. HLs are slightly more open. I would print with gamma 1.72 using studioprint and the combination of gamma and non-linear setup was the closest Iāve ever gotten to a perfect system FOR ME (Tyler Boley can probably back me up here.) I think this was a combination of using Dot Gain 19 as the linearization directive with a gamma of 1.72 but I have to look back at my notes.
What Piezography lends itself too is full continuous high-bit-depth shadow and highlight detail. Other inksets really canāt do this. So one natural way to really communicate Piezographyās unique abilities is to build a straight-line/gamma22 system and have people soft-proof on well calibrated monitors (thus simulating this straight-line gamma 22 environment). For individual people printing, this shift actually often benefits them tremendously. Suddenly the image that they are looking at that is crunched, will print with amazingly detailed HLs and Shadows that they hardly even saw in the image. So in that way, the current system reveals what is in the digital file but that is not readily seen on a normal monitor.
However, Iāve personally found that in real-world situations working for artists in a production environment, that getting a true match to ānormal-monitorā (sans soft-proof or at least somewhat sans preserve numbers) is actually a really important thing. There is significantly less translation to do between an artist, their file, their imaging abilities and brain, and my brain/system/final-tweaks. The final high-bit potential is all still there however itās then incumbent on the artist and printmaker to actually image that correct contrast ratio into the image. Having a monitor that can show that detail correctly is really key unless you go by the numbers entirely or unless youāve internalized the transition from your unique monitor setup through the print to the wall (all in your brain).
That said, the perceptual nature of the create-icc still pulls both the shadows and hls down a bit too much in my opinion. Iāve developed a way to manipulate the create-icc profile to gain the benefits of a perceptual match to the monitor while maintaining shadows and highlights (aka Studioprint). This is essentially a non-linear 1/2 perceptual system and itās super exciting because for the first time weāre able to really build for/around any previous print method along with its non-lin characteristics (for one).
This is something we are working on with Platinum printing at the moment (we are going to revolutionize alt process fyi).
cheers,
Walker
Thanks. This is actually a far more nuanced discussion than weāve had here previously, hence the blog post. I agree with all of this, at least all of it that I understand. Iām always interested in techniques to moderate the extent to which Create-ICC pulls the shadows down, on those occasions when I use that workflow. I have a couple of fairly basic techniques of my own that I use, using curves, youāll be pleased to hear, although they have their limitations.
Having made several Piezography prints for another photographer recently, I know that delivering what they want can sometimes be a challenge, especially if their monitor is a typical overly bright one a long way from calibrated for print. Which is sadly all too common. In this instance, I resorted to visiting his study and looking at the images on the screen and agreeing how the shadows would be rendered. Prints made without this information are likely not to meet expectations. Any techniques that make that process easier without losing the benefits of Piezography would be welcome.
@LGebhardt: congrats on getting a linear curve, although I remain a little puzzled about why it was so far from linear in the first place.
That bothers me too. When things are that far off it seems likely to be user error. Assuming Iām making some error Iām a bit surprised I could get the system linear. Iād still like to understand things better, and possibly fix my mistake. I tried reproducing the results with the Luster paper and I couldnāt get the droplet to accept the correction curve. I assume that is what Walker encountered, however Iām not sure what needs to be done to fix it. So Iām a little worried Iāll encounter the same issue when I move onto other papers.
Itās certainly the case that thereās some kinds of non-linearity that the droplet baulks at. Iāve had it do that on occasion when thereās a sharp kink somewhere, usually in the shadows. What Iāve done in such cases is print out the 21x4 using all candidate curves (i.e. either gloss or matte) and use the one that is closest to linearity as a starting point. Sometimes I have to use another relinearised curve as the starting point.
However if this sort of issue occurs with all or most papers, then you do have to wonder why. I donāt think that youāve got inks in the wrong order, because people who have done that get pretty wild linearisation plots, i.e. a lot more so that yours.
The exported CGATS must have the āLā letter directly above the first actual L value row. Some cgats exported from various applications have a few rows between the column title āLā and the actual values.
That is the bug.
But yeah, you do need to export as CGATS first and for-most. What I did is just recreate your L values into a cgats file.
Walker
Re the out-of-lin nature of the original print. I frankly believe this could be a result of variability in printers between 2010 (and our printer we did the curve on) and your printer. Also ink. Also if we did this on your ink-set (say neutral vs carb L differences specific to second-pass over-glossing).
Thereās a lot at play that Iām only starting to get a handle on here but time seems to be the largest factor. We keep our L values from ink-batch to ink-batch really tight but frankly there may be some un-acceptable drift there over a 5-6 year period. As a printmaker and piezography customer, I would just constantly be linearizing mainly due to printer drift and paper batches so I never relied on a ācannedā approach and never paid that much attention to those needs.
(In all the labs I ran previously I would do the same with color ICC profiling/lin too! Drift is not unique to Piezography and mostly comes down to changes in the printer.)
The result of this thread and others is that this R&D lab will be doing a full audit of our curves coupled with a validation of the current .quad droplet workflow along with a publishing of 256/step swatches/workflows for i1 and various spectro hardware/software to enable both .quad LIN and ICC profiling in the community. Top goal is to have every single user of Pieography get a better linearization than theyāve ever seen on any other print system. // that includes the nascent P2(K6) users out there. My main fear is that the current .quad droplet is not capable of creating the unique 256+ perfectly separated tones that our internal profiler (that is incredibly complicated and hard-coded into our system) can do. So that is why I keep saying I have to validate it. To validate it I have to do this on many different papers and printers.
In the next few months I could use the help of the community here who have spectrophotometers and piezo-printers that we currently donāt run in-house. I will post on that in future.
regards,
Walker
These sharp spikes (we call this āFALSESā) are actually mis-reads of the spectro. Thereās heavy math at play to fix these measurement errors so all roy software requires non-falsed measurements (actually our software does as well). Requiring a good read is the first part to building a good profile.
FYI, generally falses occur at the end of a row of patches when the spectro handles a change from gray to white. Some of the white can sneak into the reading and is often most apparent in the shadow areas where even a small amount of white pollution can effect the L value. This is most likely to happen when patches are too small as well.
regards,
Walker
Somewhere in the recent past, I think it was Jon (or perhaps Dana) who said that the publicly released curves were the result of multiple reads on multiple printers, to ensure that these curves were robust to printer variation and were representative of the wider printer population. This was in the context of why custom curves were treated with caution, as they only reflected one printer, and not one in the IJM labs either. Is that still the case?
My current Piezo printer is 5.5 years old, and Iāve kept all my linearisation plots. So I can see the amount of drift. Itās certainly a significant issue, and one that the droplet helps with enormously. However the OPās drift, if thatās what it is, is large compared with my experiences, and I thought that the Pro printers like the 3880 were supposed to be built and calibrated to a higher standard.
False readings are an issue, although I find that false measurements by my i1 are less an issue than printer variation. The averaging inherent in the 21x4 deals most variation. I have a spreadsheet that enables me to fairly quickly compare the four readouts for each patch from a 21x4, and if Iām having a struggle with a curve then I can identify whether thereās a significantly aberrant measurement. If there is, I generally find that a remeasure doesnāt solve it, I either have to fudge that measurement or reprint and remeasure.
However false readings are generally a separate issue from the droplet failing. My experience of the droplet failing is when there has been significant drift in the deep shadows, and the resulting reversal in linearity is too much for the droplet to handle. Roy has always cautioned that the linearisation maths will fail if thereās too great a reversal. (I understand the same issue occurs in QTR if youāre trying to linearise a normal QTR curve in the curve creator that has the same characteristics.)
This was my concern too. I examine the shape of the relinearised curves in curve view, and some of them can look a little strange, but then so do a few IJM curves. And I certainly canāt see any artefacts in prints, so my view is that it does a reasonable job if the non-linearity is not extreme. If by āvalidateā you mean reprint and remeasure linearity, I certainly do that. At the very least, you need to do that in order to create a soft-proofing ICC.
That would be very good. For example, there is a well-regarded master printer in this region, who became a little anti-Piezo over this issue. Having tried the system, he felt that having to buy custom curves in order to get it to work properly, in addition to the inks and carts, was unreasonable (this was some time before the arrival of the droplet). I know that thereās a lot of valuable IP in these curve creation systems, and so I understand why IJM guards it jealously, but if you want to get wider uptake, you have to help users become more self-sufficient. Which is why the droplet is such a great leap forward.
Could be variability with the printer, but I tend to doubt it. My profiles havenāt noticeably drifted over the years, and first prints from new ones produce almost exact matches to first prints made years ago (I keep them all for reference). If the ink volumes were off by as much as they were before linearizing the peizography curves I would think my color prints would have been very far off. My vote is either user error such as I have some setting wrong, or the curves I pulled down werenāt the versions I should be using.
InkjetMall Information
Why buy from us
My InkjetMall account
Contact Info