Measuring metallic inks

Cliff Crosfield of Atlas Packaging in Exeter in the UK asked a rather open-ended question of me on LinkedIn. I started to answer on LinkedIn, but then realized that I had a lot to say (go figger - like that never happens!). I decided to open the answer up to a larger audience.

“What are your thoughts on the use of spectros for measuring "metallic inks"?”

Dr. Smith turns ordinary printing inks into platinum

This is a two part question. The first is about process control and the second is about making sure the product looks right. Spoiler alert: the two parts have different answers. Spoiler alert squared: the second part doesn't have a good answer. Yet.

The topic has to do with printing with inks that contain metal flakes so as to create a metallic luster. Metallic ink is made up of flakes of metal: aluminum, copper, zinc, or silver. The metallic effect is assured when the flakes a) all lay flat, and b) cover up the entire substrate.

Flakes of metallic ink shown on left, regular ink on right

(from Rosenberg, TAGA 2001)

This blog post is about metallic inks, but some of my comments below may also apply to measurement of other sorts of processes that create a metallic luster. Many processes put ink on silvery stuff. This includes decorating (you may call it printing) beverage cans with inks that do not contain opacifers (that is, stuff that hides the metal underneath), printing CMYK inks over a silver ink (as in Color-Logic inks), and printing directly on metalized films (Mylar). Another process is laminating a printed clear film onto a metalized film, like we would see in a potato chip bag. Further afield, I am not addressing issues with metallic luster having to do with car paint, aluminum foil, or polishing up the handle of the big front door, although some of the same issues apply.

Process control

The only thing I can control on press is the amount of ink that I put on the substrate. Clearly that should be the thing that I measure. Right???!?! So, what sort of measurement best correlates with ink film thickness of a metallic ink?

Evaluation of the metallic luster of chocolate candy production

I know of three papers on this topic.

Mannig and Verdeber (of Eckart) presented a paper at TAGA in 2002 where they compared measurements of various metallic inks with the amount of ink on the substrate. Polarized and non-polarized 0/45 spectros were used for the measurements, as well as spherical instruments. Their conclusion was that measurements from polarized instruments correlated best with ink film thickness.

Mannig and Verderber - the instruments with polarization show a strong relationship

Breede and Sharma (Ryerson University) presented a paper on this subject at TAGA 2008. They compared a gloss meter (20°, 60°, and 75° incident angles) with a densitometer and a spectrophotometer (both 0/45 instruments). Their conclusion is that the gloss meter didn’t work well, but that L* or virtually any of the density channels correlates well with ink film thickness. 

The third paper was also presented at TAGA, in 2008 by Habekost and Dykopf (also of Ryerson). They compared a number of different measurement geometries: nonpolarized 0/45, polarized 0/45, spherical, and a glossmeter with three angles. They found the best correlation with the cyan channel of a polarized densitometer.

Polarized geometry seems to be the hands down favorite when it comes to measuring how much metallic ink is being put on the substrate. Why is that? I have blogged before about reflected light being a combination of bulk and specular reflectance. Of the two, we are interested in the bulk. If we want to measure the amount of ink, we want to get rid of as much of the light that reflects from the surface of the ink as possible. That suggests 0/45, and particularly polarized.

My Spectrolino, outfitted for an afternoon of measuring metallic inks at the beach

More specifically, we are interested in the percentage of the substrate that is covered with little mirrors. Mirrors have this non-intuitive property that they look black to a 0/45 or 45/0 measurement device. A first surface mirror which has been properly cleaned has a 0/45 density of over 3.0D. I recently used this property as one way to assess the design of a collection of spectrophotometers.

Why did polarization help? If you think just about the metal flakes, it shouldn't. Metals have such a high index of refraction that polarization does affect them much. Polarization helped because it reduced the effect of surface roughness of the other stiff left on the page, like varnish. 

Based on these studies and theoretical concerns, I would recommend using a polarized spectrophotometer, with either visual channel density or L*, for process control. Another recommendation based on these studies is that TAGA is a great place to look for egghead kinda papers about printing. I am not at all biased by the fact that I am the Vice President of Papers for TAGA.

Product evaluation

The second part of the question is what sort of measurement correlates best with the print buyer’s expectation from a metallic ink? 

Now, I’m not a print buyer, but I’m going to pretend for the moment to empathize with them. I’m sorry, but when the guy in the brass section gets up for his solo, it don’t mean a theen if it ain't got that sheen. I really don’t care whether the printer maintained the "correct" number of grams of ink per square meter. Metallic inks are darn expensive, and if I’m not getting the metallic luster that I want, then it isn't worth it. I’m not gonna pay big bucks to listen to Brad Paisley play jazz on a calliope even if his dog has cute earmuffs.

"I don't see what's wrong! The polarized density is well within tolerance!"

The printer may come back and say that his process is completely in control. Yesterday, a 3.2 micron thickness of ink gave him an acceptable luster. Tomorrow that should be acceptable as well. Right???!?!?

Habekost and Dykopf might beg to differ. Remember they were mentioned before? In addition to comparing measurements to ink film thickness, they also looked at visual matches. They used the corresponding metallic inks from a Pantone book as the reference, and compared sheets printed with a range of ink film thickness to these. The figure below is from their paper, with just a little bit added by me to clarify. The plot shows a lot of information, but I draw your attention to the yellow area, which represents the sheets where humans identified a good match, as contrasted against the blue area, which highlights sheets that a polarized densitometer thought were good.

Disagreement between man and machine on what constitutes an "acceptable match"

Polarized density failed miserably at predicting a visual match because it measures something other than what a human sees as being a match. Polarized density may be a good process control parameter, but in this case, it is worthless for testing conformance to the customer's real requirements. I might even say that polarized density is even worse than worthless. If a polarized densitometer were to be used to control the press, the pressman would be running to a density of about 1.6D, whereas a density of about 1.2D would have been acceptable. In addition to being a poor match, the job would have used about 30% more ink than it should have.

The thing is that so long as everything in the process is under control, process control works to provide acceptable product. In this case (the Habekost-Dykopf paper), one of the key raw materials changed. The brilliant folks at Ryerson were printing on a stock different than the stock used in the Pantone book.

Clearly we need something better than polarized density as a metric for customer acceptability. In Habekost's analysis, the spherical instrument provided the best correlation, 

Goniophotometry

It's hard for me to write a blog nowadays without mentioning goniophotometry, so, here goes...

One further technical paper is worth mentioning. This one was from Artur Rosenberg of Fogra, presented at TAGA in 2001. (Not that I am trying to push TAGA or anything.) Rosenberg used a goniospectrophotometer to measure a series of samples with varying degrees of metallic luster. His samples went from from polished metal plates to traditional inks, with a number of metallic inks in between. 

Goniospectrophotometer - the octopus of the spectrophotometer kingdom

The astute reader will recall that Rosenberg had been mentioned in my blog post about the indicatrix. I referred to this very paper.

Rosenberg pointed his goniospectrophotometer at the indicatrix in the vicinity of the gloss angle. He defined a measurement that he called “F” (standing for metallic luster) which is the product of the height and width of this specular peak – basically it is the area under the peak. In his tests, he found that this area correlates well with the perception of metallic luster.

I am still trying to get my head around this result. In Habekost's experiments, gloss meter measurements were rather poorly correlated with our perception of metallic luster. On the other hand, Rosenberg’s F is somewhat akin to the measurements that a gloss meter makes.

My own pet theory is that the width of peak is crucial for our perception of metallic luster. After all, the most metallic thing I know of, a mirror, has an indicatrix with a perfect spike at the specular angle. But, that’s further research. Maybe I’ll find a goniospectrophotometer under the Christmas tree so I can test my theories. In the mean time, I’m not sure just what to recommend to truly measure the metallic luster of metallic inks. It's pretty clear that polarized 0/45 geometry is not the choice, though.