Method to measure ink colors

[Grandad35]

2/19/08 Edit:
The changes to this post are described here.

2/20/08 Edit:
Red and Green BCI-6 inks from OEM and Formulabs were added.

3/01/08 Edit:
Red, Green, PM and PC BCI-6 inks from IS were added.

3/06/08 Edit:
Formulabs CLI-8 C/M/Y/PC/PM inks were added.

3/20/08 Edit:
A link to Canonfodders ink laydown method was added.

4/6/08 Edit:
Added Formulabs CLI-8, Ink Tec BCI-6C/M/Y/K/PM/PC and diluted versions of OEM BCI-6M

4/28/08 Edit:
Added 6:1 dilution samples of OEM BCI-6C/M and compared them to OEM BCI-6PC/PM.
Added retest samples of OEM BCI-6G and Formulabs BCI-6G.

5/20/08 Edit:
Added 10% diluted sample of OEM BCI-6M.

This is a continuation of this thread, which is itself a follow-on to this thread. The previously stated goal is to generate a catalog of CIELab colors for known inks so that we can identify an unknown ink by matching its color to known inks. A secondary goal is to make it possible to select individual ink colors from the supplier with the closest match to the OEM inks. This should give a closer color match to an OEM ink set for those who don't want to deal with profiles and are willing to run a "mixed supplier" ink set. At the very least, we would have a scientific basis for deciding which ink colors cause a color shift.

Its been quite some time since we initiated this work, but it took FAR longer to solve the ink laydown problem to our satisfaction than we (Canonfodder and I) first imagined. Youll have to take our word for it, but as a pair of engineers old-schooled in the scientific method, we beat this problem to death before we were satisfied that we would be reporting precise (repeatable) data. This post is very long, so stop reading now if you just want an executive summary there isnt one. There is a lot of information that must be conveyed to explain these tests, and there is no way to shortcut the presentation of the data.

Dont expect a recommendation for any specific ink we will present the data in what we hope is a user-friendly format and leave it up to you to draw your own conclusions. An exception was made when comparing Canons BCI-6 and CLI-8 inks, since this change has been claimed to give an improved color gamut. Since we both use Canon printers, the tests were limited to inks designed for Canon printers. There is no reason why the same test methods couldnt be applied to inks formulated for other printers, however.

Canonfodder has posted his method to laydown the ink, but Im afraid that his method has (of necessity) morphed into something that isnt easy for a casual user to replicate. To compensate for this, we have agreed to test a reasonable number of additional ink samples and to add them to this central data base.

This is a work in progress, and the testing is not yet completed. We decided to post intermediate results to gauge whether there is any interest in this work before proceeding, as it takes a lot of time to run these tests and analyze the results. When/if additional data is collected, this post and its reference links will be edited to incorporate the data and an additional post will be added to the end of this thread noting what was added/changed. If you feel that this information is useful/informative, please let us know so that we can gauge the interest.

The raw data and plots of the data are available as a spreadsheet. This spreadsheet will be updated as more samples are tested, so be sure to tell your browser to reload the spreadsheet on subsequent visits to prevent loading an old version from your browsers cache. The comparison plots are included in the spreadsheet - it is far easier to keep everything there instead of taking screen shots and posting them as jpegs in the forum. Because the raw data is included in the spreadsheet, it is also extremely easy for you to generate your own comparison plots of whatever inks you want to compare. If you dont have a program that will display an .xls file, free viewer programs are available. I have not personally used any of these, but the preinstalled version of the Open Office program on my Linux system works on this spreadsheet file:
Open Office
Google
Microsoft
NewFreeDownloads

Before starting, you might want to revisit this thread on the effects of ink dilution and this thread on the effect of how inks are combined. There are also other related threads that may also be of interest.

Open the Papers sheet (in the spreadsheet). You can see that each sample has a column of numbers associated with it. Because there is always some variation in the papers surface and how the ink is laid down, the color of each sample was measured at numerous locations (the number of sample readings taken on each sample is reported as #ofSamples). Readings are reported from 380-730 nm (the limits of human vision are usually reported as 400-700 nm). A value of 1.0 for these readings means that the paper reflects 100% of the light that strikes it at this wavelength; a value of 0.0 means that no light is reflected from the paper at this wavelength. Obviously, papers with higher readings will appear as brighter than papers with lower readings. Because the horizontal axes on the spreadsheet plots only show the row number of the data instead of the wavelength, the numbers used on the plots are given alongside the wavelengths in the chart to make it easy to find the corresponding wavelengths.

Each of these 36 numbers listed for each ink was generated from the numerous individual readings taken at each wavelength, discarding the lowest 20% and the highest 20% of the readings, then averaging the middle 60%. This was done to eliminate any readings where there might have been a hole or dark spot in the ink under the spectro (the sample size is approximately a 2 mm diameter circle). The average readings for each wavelength were then input to a program to convert the readings to an equivalent color in the CIELab color space. This color space was selected because it covers the limits of human vision and because it is the most commonly used device independent color space. The *L, *a and *b values are reported at the top of each sample.

The limit of the human eyes ability to perceive color differences is usually taken to be a Delta-E (dE) of about 1. To validate our measurement technique, the same sample was measured twice (the second set of readings was taken about 1 week after the first set) and gave a dE of only 0.5 between the two sets of readings. To test the ink applications repeatability, the same ink was laid down on two separate samples (again with about a week between the sample preparation), and gave a dE of about 2. These tests give us confidence that the test protocol works acceptably to measure the differences between inks.

The spreadsheet shows a comparison of the old Kirkland paper (made by Ilford), the new Kirkland paper (made in the US) and Kodak Ultima Photo Paper. All 3 papers give similar results at the longer wavelengths, but show significant differences at the shortest wavelengths. Specifically, the old and new Kirkland papers are quite different. Fortunately the worst of these differences are either outside of or on the far edge of our ability to see them, so the CIELab color values turn out to be quite similar. The Kodak Ultima paper shows a slight quirk at 440 nm, where it gets a reading above 1.0. Even though this might seem to be impossible, it can be caused by the addition of an Optical Brightening Agent (OBA). These agents make the paper appear whiter, but OBAs fade just like inks, so their effect usually decreases with exposure to UV light.

When starting this work last year it was decided to use the Kirkland/Ilford paper as the standard paper for our tests, and Canonfodder has set aside a large supply of this (no longer available) paper for these tests. All non-paper only tests will be made using the same box of this paper to remove the effects of the paper from our tests.

All of our spectro readings were taken with the paper placed on the white platform used to scan test patches while making a profile. Just to see the effect of the background, the Kirkland/Ilford sample was also scanned with the paper on a medium gray background. There is very little effect at the shorter wavelengths, but about 5% of the reflected light is lost to the darker background at the longer wavelengths. This points up the need to mount your photos on a white background (not the brown cardboard that comes with some picture frames) to optimize their appearance. Along the same lines, a sheet of food wrap (LDPE) was placed over some paper to see the effect. In this case, about 5% of the light was lost across the entire spectrum.

It was previously mentioned that Canonfodder had to develop a suitable technique to lay down the ink. Since this procedure takes some time and involves laying the ink down in layers, some of the ink cannot be quickly absorbed into the papers coating and can cause bronzing in some colors (especially Cyan and black). It was found that adding additional glycerin to the ink slowed the drying process, allowing the ink to be absorbed without bronzing (glycerin and certain glycols are commonly used as humectants in inkjet inks). For this reason, a sample with the same amount of glycerin (but no ink) was prepared to act as a reference. This caused an insignificant change in the CIELab values, but caused a significant increase in the reflectivity at the shortest wavelengths, reducing the differences between the old and new Kirkland papers in that area. We are not sure what this means from a practical standpoint, if anything. Just to satisfy our curiosity, a similar sample was made with glycerin on the Kirkland/US paper with glycerin, the spectra on these two papers are much closer at the shorter wavelengths.

From ink usage data gathered from several sources, it has been determined that a total of about 0.6 CCs of ink is used to print an average 8x10 photo. This works out to an average of 11.6 microns of total ink, or about 4 microns for each main color. Our test lays down about 7.2 microns of a single ink, which gives a medium-dark coverage for a single ink. The ink density was determined by a series of visual tests at different laydown rates using the individual C/M/Y inks, and was only calculated once we were satisfied with the appearance. All of the ink tests presented here were made at this 7.2 micron laydown rate. Changing the laydown rate would change the absolute CIELab values for the various inks, but the relative comparisons between inks should still be the same. As such, these results are only meaningful when comparing inks applied according to Canonfodders procedure our CIELab values are meaningless when compared to samples prepared in any other manner.

An inkjet printer lays down dots of full strength ink, leaving unprinted white areas between the dots on light areas of the photo. It does NOT spread out the ink uniformly over the entire surface as we did for our tests. The ideal concept that we started with was to develop a way to have a printer deposit a single ink in a controlled manner, but we could not find a practical way to make this happen. Using a dot diameter of 60 microns and an ink drop volume of 2 pl (2000 cubic microns) results in an average ink thickness of 0.7 microns for each drop of ink deposited by itself. This means that (on average) 10 overlapping drops of ink have to be deposited to get the same coverage as we used, so using a uniform application of ink to prepare the samples isnt as bad as might initially be supposed. My i9900 is rated at 2400x4800 dpi (10x5 microns/dot) - if a 2 pl ink drop was deposited at each and every possible position, this works out to a maximum 36 micron layer of ink (for each color). The test value of 7.2 microns appears to be a reasonable value from every angle that we investigated.

Some of the early comments in this thread indicated a skepticism about the ability to compare inks using this method. Hopefully, including the raw spectro data in the spreadsheet format will show far more detail than just the CIELab values and also make it extremely easy for others to make their own comparison plots from the raw data. To add an ink to an existing graph, just drag over the spectrum values (rows 7-42 in the selected column), hit "Ctrl-C" (copy), click on the desired graph (even if the graph is on a different page), then hit "Ctrl-V" (paste). To create a new graph from scratch, just use one of the spreadsheet's many graphing options.

Two different inks could have the same CIELab values but have different spectral plots, making them appear differently when combined with other inks or when viewed under different lighting conditions (metamerism). However, if two inks have the same spectral response, they should look the same when combined with other inks or viewed under different lighting conditions. Plots of the full spectrum will also show such things as whether one ink is just a more concentrated version of the other.

These tests do NOT look at the fade resistance or tendency of a given ink to clog - only their color. We know from previous tests that Canons OEM inks have superior fading resistance to any of the 3rd party inks that we tested.

Along similar lines, the OEM inks have always been considered to be the gold standard that all other inks should try to match when it comes to color. This may well be true, given that Canon has spent a lot of money developing their ink set and has programmed the printers firmware specifically around the characteristics of their ink set. However, we have no proof that a differently balanced ink set may not also give an acceptable color gamut. A custom printer profile can certainly compensate for minor differences in the ink colors (If you are fussy about your colors, definitely plan on getting a custom printer profile), as long as the printers firmware plays nice with the colors in the 3rd party ink set tests of this sort are well beyond the scope of this work. For this reason, we will report the dE between inks, but will not make any judgment as to which ink set is better.

At least as important as the color values that we measured is how repeatable the colors are from bottle to bottle. We have not seen any reports of inconsistency from any of the major 3rd party ink manufacturers (e.g. Image Specialists, Formulabs, OCP).

In case you missed the link above, you can find the spreadsheet here.

Lets start with the yellow inks (select the Yellow sheet on the bottom of the spreadsheet), since yellow has the simplest spectrum of C/M/Y inks. We see:
*The Formulabs BCI-6Y and OEM BCI-6Y match fairly well across the spectrum (dE=2.1).
*The Formulabs CLI-8Y differs from the OEM CLI-8Y (dE=7.5).
*The Formulabs BCI-6Y differs from the Formulabs CLI-8Y (dE=5.2)
*The IS BCI-6Y differs from the OEM BCI-6Y (dE=11.8). Note that the IS CLI-8Y is the same ink as IS BCI-6Y.
*The MIS CLI-8 differs from the OEM CLI-8 (dE=13.1), but closely matches IS BCI-6Y (dE=1.2).
*The Ink Tec BCI-6Y differs from the OEM BCI-6Y (dE=4.5).
*The OEM BCI-6Y and OEM CLI-8Y are almost a perfect match at the shorter wavelengths, but the OEM CLI-8Y reflects MORE light at the longer wavelengths than the paper alone! The CIELab values for these two inks are a very good match (dE=0.9). It would be easy to postulate that the behavior of the OEM CLI-8Y should give slightly brighter yellows/reds than the OEM BCI-6Y, and is consistent with Canons claims.
*It has been rumored that MIS supplies IS inks, and the IS website lists MIS as their North American distributor]. The MIS CLI-8Y is close to the IS CLI-8Y(also BCI-6Y), including the little kink at 580 nm.
*Our initial test of the MIS CLI-8Y (CLI-8Y_MIS_Orig on the spreadsheet) looked strange at the longer wavelengths, so a retest was run (CLI-8Y_MIS on the spreadsheet), and the retest data is included in this analysis. Canonfodder is very meticulous in his ink application procedures, so we are fairly certain that the problem wasnt in contamination during the application of the ink, but rather that the Orig ink sample was pulled out of a 2 oz. bottle that was almost completely empty. It is possible that there was some foreign material in the dregs that affected the color of that sample. The retest sample was drawn from a nearly full refilled cart. Obviously, we will be very careful about how the samples are obtained for future tests.

Select the Cyan sheet:
*The OEM BCI-6C and Formulabs BCI-6C match fairly well across the spectrum (dE=1.1).
*The Formulabs CLI-8C differs from the OEM CLI-8C (dE=12.8).
*The IS BCI-6C differs from the OEM BCI-6C (dE=10.8 with the peak shifted slightly to the right). Note that the IS CLI-8C is the same as IS BCI-6C.
*The MIS CLI-8C differs from the OEM CLI-8C (dE=8.7 with the peak shifted slightly to the right).
*The Ink Tec BCI-6C differs from the OEM BCI-6C (dE=16.1).
*The OEM BCI-6C and OEM CLI-8C have a similar spectral response, but the OEM CLI-8C behaves like a slightly diluted version of the OEM BCI-6C (dE=3.0), with its peak at exactly the same wavelength for both inks. This OEM CLI-8C isnt simply a diluted version of the OEM BCI-6C, however, since it absorbs more light in the 570-630 nm band. It appears to be an improved formulation, and these characteristics should allow more saturated cyan colors.

Select the Magenta sheet:
*The magenta inks are apparently the most difficult to formulate, since they have to allow light to pass at both ends of the spectrum while absorbing everything in the middle wavelengths (they also fade the fastest).
*The Formulabs BCI-6M differs from the OEM BCI-6M (dE=5.7).
*The Formulabs CLI-8M differs from the OEM CLI-8M (dE=3.7), but is closer than the BCI-6.
*The shift in the color from the Formulabs BCI-6 to CLI-8 is similar to the shift from OEM BCI-6 to CLI-8.
*The IS BCI-6M appears to provide a good match to the OEM spectrum, but the dE is larger (dE=9.6).
*The Ink Tec BCI-6M is close to the OEM BCI-6M (dE=2.2).
*It is obvious that Canon made a major change to their magenta formulation on the OEM CLI-8M, with a dE of 17.7 between the BCI-6 and CLI-8 inks. Note that the *L and *a CIELab values are virtually identical, but that the *b value has a large negative shift with the new ink, shifting toward blue and away from yellow.
*The MIS CLI-8M differs from the OEM CLI-8M (dE=7.6).
*The IS CLI-8M and MIS CLI-8M are fairly close (dE=3.0). The spectra are very similar and the differences could be explained by batch-to-batch dilution differences caused by how well the ink was remixed before the consumer sized bottles were filled. The MIS CLI-8M could be the same as IS CLI-8M.
*The IS BCI-6M and IS CLI-8M inks shows a smaller change in their formulation than is seen with the OEM inks (dE=5.4 for IS vs. dE=17.7 for Canon).
*It may not mean anything, but the Canon spectra are both nice and smooth, while the 3rd party spectra have kinks (the Canon plots appear to be more natural). As with the OEM CLI-8 yellow and cyan inks, the spectrum of the OEM CLI-8M should give an improved color range over the OEM BCI-6M. Canons claim of an improved color gamut with the CLI-8 inks appears to have a basis in fact.

Select the Black sheet:
*The Formulabs BCI-6K is weaker than the OEM BCI-6K at the shortest and longest wavelengths, but it is slightly stronger in the middle (dE=4.5).
*The Formulabs CLI-8K is stronger than the OEM CLI-8K at the shortest wavelengths, but is weaker at the longest wavelengths (dE=4.0).
*The Formulabs CLI-8K is darker that the Formulabs BCI-6K and appears to be a more neutral color.
*The OEM CLI-8K appears to be a major improvement over the OEM BCI-6K, especially at the longer wavelengths, as it absorbs more light at all wavelengths (dE=8.3).
*The IS CLI-8K is a closer match to the OEM BCI-6K (dE=4.7) than it is to the OEM CLI-8K (dE=7.4).
*The Ink Tec BCI-6K is a reasonable match to the OEM BCI-6K (dE=3.4).
*Remember that these inks were applied at a medium-dark rate, so a full application of any of these inks by a printer could easily result in less reflected light than is shown here.
*During our fading tests, it was found that the blacks often faded with a red cast. Since the effect of fading is similar to the effect of diluting an ink, it is clear that these inks will tend to allow more of the longer wavelengths through as they fade. This appears to be one of the big improvement with the OEM CLI-8K it should be more fade resistant even if no other changes were made to the chemistry.

Select the PM (photo magenta) sheet:
*The OEM BCI-6PM differs from the OEM BCI-6M by a very large number (dE=36.6). Such comparisons are meaningless, and the spectra need to be analyzed to see the differences between M and PM inks.
*The spectrum of the OEM BCI-6PM has about the same shape as the OEM BCI-6M, but appears to have about 20-25% of the strength of the Magenta.
*The Formulabs BCI-6PM also seems to be a 20-25% strength version of the Formulabs BCI-6M.
*The Formulabs BCI-6PM differs from the OEM BCI-6PM by about the same amount as the magenta inks (dE=6.0), even though the spectra appear fairly different.
*The IS BCI-6PM matches the OEM BCI-6PM (dE=3.3) better than the full strength magenta (dE=9.6).
*The Ink Tec BCI-6PM differs from the OEM BCI-6PM (dE=4.9).

Select the Diluted sheet:
*This data shows OEM BCI-6C and OEM BCI-6PC, as well as a 17% diluted versions of the OEM BCI-6C.
*The 17% diluted sample is a very good match for the PC (dE=4.8).
*This data shows the OEM BCI-6M and OEM BCI-6PM, as well as 50%, 25%, 17% and 10% diluted versions of the OEM BCI-6M.
*The PM ink has about the same spectral response as the diluted Magenta. The 10% diluted sample is a very good match for the PM (dE=2.2).
*It is small wonder why PM and PC inks have the highest usage rates. Printers that use small drops of M and C to simulate PM and PC should have significantly lower ink usage.
*While diluting full strength M and C might seem like a good way to save money on PM and PC inks, simply diluting full strength inks with water would result in the wrong viscosity and surface tension, as well as lower concentrations of other additives. It is not recommended.

Select the PC (photo cyan) sheet:
*As with the Photo Magenta inks, the OEM and Formulabs PC inks appear to be 20-25% strength versions of their cyan inks.
*The Formulabs BCI-6PC is about 25% stronger than the OEM BCI-6PC at the longer wavelengths, but about the same at the shorter wavelengths (dE=9.6 for the PC inks vs. dE=1.1 for the Cyan inks).
*The IS BCI-6PC matches the OEM BCI-6PC (dE=4.6) better than the full strength cyan (dE=10.8).
*The Ink Tec BCI-6PC differs from the OEM BCI-6PC (dE=6.4).

Select the Red sheet:
*The Formulabs BCI-6/CLI-8R (the same ink is used for both) is close to the OEM BCI-6R (dE=1.6), even though the OEM ink appears to be much darker in the cart. This proves our suspicion that the liquid ink color is not a reliable predictor of the color that appear when the ink is applied to paper.
*The IS BCI-6R has a different spectrum than the OEM BCI-6R (dE=23.2).

Select the Green sheet:
*The OEM BCI-6G was retested, and the 2 samples fell within a dE of 3.2. The data for the first test was used for the plots (both sets of data were included for reference in case you want to compare them).
*The Formulabs BCI-6G was also retested (because the first comparison with the OEM BCI-6G looked strange). The retest matched the OEM ink more closely, and was used in the comparison. I dont know for sure what happened with the first test, but it appeared to be much stronger than the retest. Perhaps I didnt shake the bulk ink container before pulling the first sample and got a concentrated sample?
*The Formulabs BCI-6/CLI-8G the same ink is used for both) differs from the OEM BCI-6G (dE=7.2). Both the samples and the spectra show the same thing the Formulabs ink is stronger than the OEM ink. The spectra have similar shapes and the peaks and bends all occur at the same wavelengths (see the IS BCI-6G for an example of different inks) - the differences are consistent with a simple difference in concentration.
*The IS BCI-6G has a different spectrum than the OEM BCI-6G (dE=24.6).

The following is based more on conjecture than on knowledge. I have read a great deal on the subject of color, but have never seen limitations in printed gamuts explained in this way. Hopefully, someone with real knowledge in this area will chime in. The OEM Inks sheet in the spreadsheet shows the C/M/Y OEM inks plotted together. Unlike an RGB monitor, which creates light in an (additive color system), inks absorb light in a (subtractive color system). As such, it is important to consider how well an ink can absorb as much light as possible in a certain band of wavelengths while passing as much light as possible in the other wavelengths. Especially note how each of the three C/M/Y inks has a band of wavelengths where they are almost completely opaque this is why these 3 inks can generate an almost black when combined.

The Red can be seen to be similar to the yellow, except that it allows more of the middle wavelengths to pass. To properly utilize this capability the printers firmware (its RIP) must be programmed appropriately, but it can be seen that the printers color gamut can be extended. The Green behaves more like the cyan - it tends to pass only a range of wavelengths in the middle of the spectrum. As with the red, the firmware must be programmed to take this capability into account, but it allows a wider color gamut. Green can also be printed by combining full strength yellow and low strength cyan, but the center of the green peak will be shifted toward the shorter wavelengths.

The spectra of the yellow inks are almost perfect - they are almost completely transparent at the longer wavelengths, but very opaque at the shortest wavelengths. The cyan and magenta are also very opaque in the longest and middle wavelengths, but they are only partially transparent at the shorter wavelengths. This introduces a limitation in the range of colors that can be printed. If it is necessary to absorb all of red light (using cyan), much of the shorter wavelength light will also be lost; to pass more of the shorter wavelengths, less cyan must be used, passing more of the longer wavelengths. This is why the CLI-8 magenta is an improvement the middle wavelengths can still be absorbed while losing less of the shorter wavelengths. Note that the left peak on the CLI-8M is now almost as high as the CLI-8C; it is only about as high on the BCI-6 inks. If Cyan and Magenta inks could be developed that had the same characteristics as the yellow ink (transparent outside of their opaque band), a wider range of colors could be printed, greatly expanding a printers color gamut.

To round out this discussion, it is also instructive to look at how our eyes perceive color (see the spectra part way down) and how monitors generate colors. The Monitor Color sheet shows the spectro readings that were taken on my calibrated CRT (conventional phosphors) and calibrated LCD display when displaying the additive and subtractive color charts. The plots on the right of the data show the spectra of the light emitted for each color (the vertical scale is relative, and the specific numbers have no special meaning in this case). The green and blue emitted by the conventional CRT are at about the same wavelengths as our eyes sensitivity, and have a similar normal distribution. The red, however, is shifted to the right and is made up of two separate narrow peaks of light instead of a single smooth distribution. The light from the LCD monitor is quite different from the CRT, with the spectra from all 3 colors appearing as jagged lines. In each case, it can be seen that the R+G+B lines are the numerical sum of the light from the 3 separate colors at each wavelength, just as would be expected in an additive system. The C/M/Y data is also presented for reference, showing how a monitor simulates the subtractive colors using combinations of the additive colors (at least as closely as is possible due to the limitations of the R/G/B spectra). Even though the spectra for C/M/Y are very different from the spectra measured on the inks, our eyes still perceive them to be similar colors.

The gamuts of the CRT and LCD were compared, and are shown here. This is only a single view of a 3-D image, but it demonstrates that this CRT can generate some colors that this LCD cannot, while this LCD can generate other colors that this CRT cannot. For example, note that the CRT is superior in the blues this corresponds to the peak of the CRTs blue spectrum being located to the left of the peak of the LCDs blue spectrum. The gamut of the CRT fits the sRGB color space more closely than the LCD, and may be the reason why many professionals prefer high end CRTs for critical color work (this CRT is NOT a high end unit).

If this discussion hasnt given you a headache by now, you havent been paying attention. If you have developed a raging headache, you are beginning to see just how difficult it is to generate the same perceived colors on different devices and why various devices can have very different color gamuts. This is one of those subjects where the more that you study and learn, the more confusing it seems to become. It has certainly given me a new appreciation for just how much technology goes into developing an inkjet printer and formulating an inks color. Even with the limitations in their color gamuts, its amazing that any of this stuff works as well as it does.

 

[stratman]

This ia a fascinating topic that gets way above my payscale, but I am enjoying the read.

I don't know if it would be of help, but some rough estimates of the proportion and type of some of the ingredients of Canon inks can be gleaned from the MSDS sheets: http://www.canon.com.au/support/msds.html.

I am a bit confused as to which aftermarket ink mimics the OEMs best. Even more elemental, were your test aftermarket inks all comparing apples to apples? I mean, did you use the aftermarket inks specifically formulated for both BCI and CLI ink types? For instance, There does not appear to be a CLI-compatable IS ink for CYAN, though there are CLI-compatable IS inks for all the other dye based colors. Since I have a CLI ink printer, and I am using IS ink, I would have liked to have seen data about the all CLI-compatable IS inks. Will you be adding more data about CLI-compatable IS inks or is the CYAN IS ink the same in both BCI and CLI compatables as it is for YELLOW?

I couldn't find the identifying code for each aftermarket ink you had in your post. Please provide the specific identification alphanumerics for each aftermarket ink used so that we can purchase the precise ink we feel best matches. For instance, my IS ink is coded as:

Cyan - C2032D
Yellow - C0797D
Magenta - C6121D
Black (dye) - C1109D

Thank you for the indepth analysis. Look forward to more information in the future.

 

[Grandad35]

Stratman,

The link to the MSDS sheets doesnt work for me. For the purposes of this work, we arent really looking at what is used to formulate the inks only their colors.

It isnt our goal to say which inks provide the best color match, only to provide the raw data so that others can come to their own conclusions. I think that it is clear that none of the 3rd party ink sets that we have tested so far closely matches all of the OEM colors. As we have also stated, However, we have no proof that a differently balanced ink set may not also give an acceptable color gamut. It is beyond the scope of our work to look at the color gamuts of each complete set of inks when printed by a properly profiled printer.

I guess that I should have mentioned that each of these inks is sold as being formulated for the stated cart type. The sources for each ink set were as follows:
OEM BCI-6 - Supplied by me by blowing out ink from OEM carts.
OEM CLI-8 - Supplied by Canonfodder by blowing out ink from OEM carts.
Formulabs BCI-6 - Supplied by me from my bulk ink containers. I dont know the ink numbers, but they are these inks.
MIS CLI-8 supplied by Canonfodder from his bulk ink containers.
IS BCI-6 and IS CLI-8 - supplied by Mikling from his bulk ink containers.

We have NOT completed our testing (e.g. PM/PC/R/G). I am not sure if the IS CLI-8C is the same as the IS BCI-6C or if Canonfodder has just not made that sample yet. Perhaps they can chime in on that color, as well as on the specifics of any ink identification numbers that they have for their samples.

 

[canonfodder]

Stratman and Grandad35,

The ink company Image Specialists, Inc., located at www.image-specialists.com gives the following information on BCI-6 and CLI-8 dye-base ink equivalents which they supply. Note that both cyan and yellow which they supply are the same ink product for BCI-6 and CLI-8 cartridges, while both magenta and dye black have different part numbers for these two cartridge families.

BCI-6C and CLI-8C are both I.S. # WJ2032.

BCI-6M is I.S. # WJ6053.

CLI-8M is I.S. # WJ6121.

BCI-6Y and CLI-8Y are both I.S. # WJ797.

BCI-6BK is I.S. # WJ1008.

CLI-8BK is I.S. # WJ1109.

Stratman, I presume your supplier of Image Specialists inks changes the prefix from WJ to C. The numbers match otherwise.

 

[mikling]

I change the prefixes so that a particular ink is quickly identified for a certain brand of printer. C would denote Canon and E would denote Epson etc. The prefix of WJ otherwise would serve no purpose external to IS as that is used for all the inks by Image Specialists . As to the suffix, D is used to denote dyebased as opposed to a suffix of P to denote Pigment. This allows one to quickly identify what brand of printer a particular ink formulation is targeted for and its composition.

 

[stratman]

Stratman, I presume your supplier of Image Specialists inks changes the prefix from WJ to C. The numbers match otherwise.

You are correct. My inks are from Mikling of precisioncolors.com.

Thanks to Grandad35, Cannonfodder and Mikling for the clarifications.

 

[Grandad35]

While looking up the catalog numbers for the Formulabs inks, I came across this article on how some resellers say that they supply ink from Company A, but dont say that it is being sold for a different cart series than the ink was formulated for. This could explain some of the strange reports that we got a year or two ago about how poor the color was from some prefilled carts, even though they were supposed to be filled with a name brand ink. It makes a good read. Its easy to see why a 4 gallon/color minimum order would make it tempting for a small reseller to try to sell one ink for multiple applications. This link gives an idea of the number of inks that a full-line reseller would have to stock to service every printer.

The Formulabs and IS catalog numbers have been added to the spreadsheet. If you want the colors of the bulk inks/prefilled carts that you purchase to match the data in the spreadsheet, be sure to specify the appropriate suppliers catalog numbers when you order.

 

[Tin Ho]

Grandad, the Sensient Technology catalog shows something that makes me wonder a bit. It shows that MP830 uses PGI-40 and CLI-41 whcih are wrong. It in fact uses PGI-5 and CLI-8s. It shows that PGI-40 and PGI-5, as well as CLI-41 and CLI-8 are different inks. But Canon has always advertises that CLI-8 and CLI-40 are both Chromalife100 ink. Is there any difference between the ink in CLI-8 and in CLI-41 regardless they are all Chromalife100 ink according to Canon? If they are indeed slightly different why would Canon want to make them different? Is there any real technical reasons to make them different? Frankly there is a lot of myth and commercial secrets in inkjet inks. Can we really believe that there are really so many different inks according to the catalog from Sensient Technologies? How can you tell that it's not a marketing hype?

HP, Canon and Epson all claim to have advanced R&D in the inks they make. They all claim some high tech ingredients in their ink for specific advanced features and they all have tons of patents to protect their inks from being cloned. From this point I can believe these OEM companies do have a lot of different inks, each differs in the special ingredient for specific features. I wonder if Sensient Technology also comes out with same number of different inks so that each one is specifically compatible with one specific OEM ink. If so what are the differences in their inks? They can not use any patented formula to make their ink compatible. I am not convinced that they really have so many different inks, each one is formulated differently in order to be fully compatible with each different OEM ink that is different in specific additional ingredient for specific advanced feature.

I also highly suspect that Sensient Technologies makes OEM inks for any of HP, Canon and Epson. If it does they will not be allowed by HP, Canon or Epson to make compatible inks to compete with the ink they make for these three OEM companies. If they allow Sensient Technologies to do that. Sensient Technologies will have very best inks to beat OEM inks in no time. I don't think HP, Canon or Epson would be so foolish to allow it to happen. However, I do believe Sensient Technology makes a lot of Formulabs inks in China. China is the largest ink market for any 3rd party ink manufacturer to miss. So it is true that a lot of prefilled after market ink cartridges are filled with Formulabs ink in China. I don't think the ink is made in the USA then shipped to China for the market there. China also has a lot of ink manufacturers itself. I tend to believe HP makes its ink in the USA and Epson and Canon both make their ink in Japan.

To me there is a lot of hype and myth among after market and OEM inks. We need to take marketing pitches into consideration and evaluate price/performance when buying bulk ink for our printers. It is very easy to add a lot of part numbers and play games on their prices. Sometimes I do wonder why some generic inks are selling well. IMS ink from Costco is one example. Despite being generic it apparently works OK for a lot of people.

 

[Grandad35]

You bring up several valid points that demonstrate how little we really know about the ink used in our printers. During the course of this project and the fading project, I have had the opportunity to communicate with a few resellers. My sense of these communications is that the resellers that I have had contact with are honorable business people who have been burned in the past by mis-information and by people who have misapplied their products. They have been invited to share their information in this forum, but every one of them is afraid that they might say something that could be misconstrued and blown out of proportion. Mikling obviously has a lot of knowledge in this area and would be able to provide a lot of insight into these types of questions, but I completely understand that he may not be able to share this knowledge in this area with us. This is one of the reasons why we undertook this project to scientifically measure the ink colors with an impartial instrument instead of trusting statements like I have a great eye for color, and Sams ink gives great colors. As you read about this subject, it quickly becomes clear that humans dont all possess the same ability to judge colors and that the population forms a continuum from completely color blind to having superior color sensitivity.

This thread is a great place to discuss these types of issues, and I hope that anyone with knowledge in this area will share their knowledge with us. Lacking such knowledgeable sources for time being, Well be limited to just giving opinions.

The Sensient Technology catalog shows something that makes me wonder a bit. It shows that MP830 uses PGI-40 and CLI-41 which are wrong. It in fact uses PGI-5 and CLI-8s. It shows that PGI-40 and PGI-5, as well as CLI-41 and CLI-8 are different inks. But Canon has always advertises that CLI-8 and CLI-40 are both Chromalife100 ink. Is there any difference between the ink in CLI-8 and in CLI-41 regardless they are all Chromalife100 ink according to Canon? If they are indeed slightly different why would Canon want to make them different? Is there any real technical reasons to make them different? Frankly there is a lot of myth and commercial secrets in inkjet inks. Can we really believe that there are really so many different inks according to the catalog from Sensient Technologies? How can you tell that it's not a marketing hype?

We cant really be sure about this, but I doubt that it is all hype. Chromalife100 doesnt refer to a specific set of ink colors, but a system of printer/ink/paper, so it is possible for more than one set of inks to have this designation. I have no experience with these carts and this is only conjecture, but is it possible that a different nozzle configuration might require a different additive package in the various carts? If only C/M/Y dye inks are available for printing photos (CLI-41), it is also possible that the ink colors are optimized (in the CLI-41) to compensate for the lack of a dye black.

Even within one printer family using the same basic technology, it has been shown that the CLI-8 inks are different than the BCI-6 inks, and that they offer an improvement in the color gamut (and maybe in longevity). Canon cant modify the formulation for their BCI-6 inks to incorporate these types of improvements (imagine the uproar this would cause with owners of older printers using those carts), so they have to wait until they introduce a new line of carts to make any improvements in their inks. Their cart volume is so high that it isnt a big problem for them, and they arent exactly concerned about the inventory problems that this might cause for 3rd party resellers (or retail stores, for that matter). The Sensient chart shows that some inks ARE used in multiple carts (e.g. C/M/Y in BCI-3/BCI-3e/BCI-6/BCI-7/BCI-15 are the same).

I wonder if Sensient Technology also comes out with same number of different inks so that each one is specifically compatible with one specific OEM ink. If so what are the differences in their inks? They cannot use any patented formula to make their ink compatible. I am not convinced that they really have so many different inks, each one is formulated differently in order to be fully compatible with each different OEM ink that is different in specific additional ingredient for specific advanced feature.

There is no way for us to know the answer to this question, and it is only by running our own tests that we can verify that a specific ink set matches the OEM carts in our printers in terms of color and longevity.

I also highly suspect that Sensient Technologies makes OEM inks for any of HP, Canon and Epson. If it does they will not be allowed by HP, Canon or Epson to make compatible inks to compete with the ink they make for these three OEM companies. If they allow Sensient Technologies to do that. Sensient Technologies will have very best inks to beat OEM inks in no time. I don't think HP, Canon or Epson would be so foolish to allow it to happen. However, I do believe Sensient Technology makes a lot of Formulabs inks in China. China is the largest ink market for any 3rd party ink manufacturer to miss. So it is true that a lot of prefilled after market ink cartridges are filled with Formulabs ink in China. I don't think the ink is made in the USA then shipped to China for the market there. China also has a lot of ink manufacturers itself. I tend to believe HP makes its ink in the USA and Epson and Canon both make their ink in Japan.

*This article confirms that the major ink suppliers also make ink for the OEMs. I know an engineer who used to work for a company that made equipment used in making ink. He was told the same thing, but that they were free to make inks using their own formulations as long as they didnt use the OEMs ink technology or patents. Its obvious from the fading results and the differences in the colors that the formulations are different.
*He also said that the facility that he was at (somewhere in the Americas he couldnt identify the company or the location because of a secrecy agreement) told him that they made some of the inks for Canon, Epson and HP.
*According to the Sensient website, their inkjet ink manufacturing facility is in Mexico.
*According to the IS website, their ink is manufactured in Long Island (New York).
*The OCP website says Inks made in Germany.
*Any of these companies could obviously opeatate a Chinese subsidiary under a different name, but they couldnt sell that ink under the expensive name (and there is no evidence that any of these three suppliers are doing this).
*The previously mentioned article also explains how some of the Chinese carts could be filled with Formulabs ink, but still have the wrong colors because the ink that they are using was designed for a different cart.

To me there is a lot of hype and myth among after market and OEM inks. We need to take marketing pitches into consideration and evaluate price/performance when buying bulk ink for our printers. It is very easy to add a lot of part numbers and play games on their prices. Sometimes I do wonder why some generic inks are selling well. IMS ink from Costco is one example. Despite being generic it apparently works OK for a lot of people.

Read this thread. How many I care about longevity responses did you count? The vast majority of people who buy 3rd party carts and bulk ink dont care about longevity or color just price. The Costco product is apparently good enough for them. If people knew that far better products were available at the same price they might be interested, but how would this information be conveyed to them? This is the most active forum on this subject and it currently has 2800 members a drop in the bucket compared to the overall market size for 3rd party carts and bulk inks.

 

[Tin Ho]

Grandad, thanks for the long post full of excellent information. It takes me a while to digest it. On the subject of ink longevity I think there are ink that will fade badly in a month or so. I once chatted with my uncle in Los Angeles. He said he could buy BCI-6 prefilled in LA's convention center for $0.70 a pop. But the colors will fade in a month. Guess what. He said he did not care. Well, it is probably true. People send printers to landfill when ink runs out. If then can find $0.70 cartridges they will hold on to the printer a little longer. These people do not print photos. They probably print a few sheets of something a month. The printer with the original cartridges will last more than a year for them. With cheap compatibles they can hold on for another year until the print head dies then it goes to landfill. I doubt the printer ever finishes printing a realm of paper.