i1Pro good repeatibility for 2000+ patch targets

[Jalok]

I recently purchased an i1Pro Xtreme bundle, upgraded it to i1Profiler Publish software, and before exploring its 2nd iteration profiling, I decided to investigate the spectrophotometer repeatability varying targets between scrambled and non-scrambled patches and between 7.3 and 10mm width printed patches. For the first moment, I printed four 400-patch targets on normal paper, waited overnight to dry and scanned each of them twice, with a moderate to fast movement speed (around 2 seconds per row) through the scanning ruler.

Comparing each measurement to the other of the same target, i1Profiler Data Analysis gave me the following results:

-- 10mm patch width and scrambled patches:

Average DeltaE (deltaE2000) results [Measured (Patches)]

Total: 0.10
Lowest 90%: 0.09 (360)
Highest 10%: 0.22 (40)

Standard deviation results
Total: 0.06
Lowest 90%: 0.04 (360)
Highest 10%: 0.04 (40)

Maximum results
Max DeltaE all patches: 0.33
Max DeltaE in the lowest 90%: 0.18 (360)


-- 7.3mm patch width and scrambled patches:

Average DeltaE results
Total: 0.26
Lowest 90%: 0.20 (360)
Highest 10%: 0.79 (40)

Standard deviation results
Total: 0.25
Lowest 90%: 0.11 (360)
Highest 10%: 0.44 (40)

Maximum results
Max DeltaE all patches: 2.19
Max DeltaE in the lowest 90%: 0.49 (360)
 

-- 10mm patch width and non-scrambled patches:

Average DeltaE results
Total: 0.11
Lowest 90%: 0.10 (360)
Highest 10%: 0.27 (40)

Standard deviation results
Total: 0.07
Lowest 90%: 0.05 (360)
Highest 10%: 0.06 (40)

Maximum results
Max DeltaE all patches: 0.47
Max DeltaE in the lowest 90%: 0.21 (360)

 
-- 7.3mm patch width and non-scrambled patches:

Average DeltaE results
Total: 0.14
Lowest 90%: 0.11 (360)
Highest 10%: 0.45 (40)

Standard deviation results
Total: 0.13
Lowest 90%: 0.07 (360)
Highest 10%: 0.12 (40)

Maximum results
Max DeltaE all patches: 0.86
Max DeltaE in the lowest 90%: 0.32 (360)

--

Since the best statistics were reached with wider patches for the scrambled target and non-scrambled target for the narrower patches, I had made once again the first and the forth measurements with a slower scanning speed through the ruler (around 4 seconds per row), and got the following results:

-- 10mm patch width and scrambled patches:

Average DeltaE results
Total: 0.12
Lowest 90%: 0.11 (360)
Highest 10%: 0.25 (40)

Standard deviation results
Total: 0.06
Lowest 90%: 0.05 (360)
Highest 10%: 0.04 (40)

Maximum results
Max DeltaE all patches: 0.40
Max DeltaE in the lowest 90%: 0.21 (360)
 

-- 7.3mm patch width and non-scrambled patches:

Average DeltaE (DeltaE2000) results
Total: 0.12
Lowest 90%: 0.10 (360)
Highest 10%: 0.32 (40)

Standard deviation results
Total: 0.09
Lowest 90%: 0.06 (360)
Highest 10%: 0.07 (40)

Maximum results
Max DeltaE all patches: 0.53
Max DeltaE in the lowest 90%: 0.25 (360)

As someone could see, these two last measurements gave me similar numbers. As it became evident scanning speed really matters, I do not mind scanning them slower for the sake of more precision and better repeatability. But it would be great if I could maintain the 7.3mm width patches as they produce much less printed targets (for huge number of patches) and less scanning work too. But again, I do not want to loose any necessary precision of repeatability, but do not mind loosing unnecessary repeatability. Does anyone could tell me if the last measurement (7.3mm non-scrambled) has sufficient good repeatability for 2000-4000 patch targets? In other words, if their average deltaE(2000) are smaller than the small patch-to-patch differences in those thousand-numbered targets?

[aaronchan]

The iO or iSis would be something that will "only" do spot measuring with a extreme fast speed.
I use my Z3200 to read most of my targets now since it does the same thing as well.

[Ernst Dinkla]

The iO or iSis would be something that will "only" do spot measuring with a extreme fast speed.
I use my Z3200 to read most of my targets now since it does the same thing as well.

I have been looking into that too. The Z3200 Color Center target import and measurement output is however quite unpredictable, getting it to work is one thing, saving the measured data on the computer another thing. With APS it goes better but that one only exports spectral data. BTW. the spectrometer doesn't make 380, 390 and 410 Nm spectral readings, 400 Nm is included and at the red end it skips the 720 + 730 Nm wavelengths. Is that similar to the UV cut iSis?

met vriendelijke groeten, Ernst

New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[Bart_van_der_Wolf]

[...]
As someone could see, these two last measurements gave me similar numbers. As it became evident scanning speed really matters, I do not mind scanning them slower for the sake of more precision and better repeatability. But it would be great if I could maintain the 7.3mm width patches as they produce much less printed targets (for huge number of patches) and less scanning work too. But again, I do not want to loose any necessary precision of repeatability, but do not mind loosing unnecessary repeatability. Does anyone could tell me if the last measurement (7.3mm non-scrambled) has sufficient good repeatability for 2000-4000 patch targets? In other words, if their average deltaE(2000) are smaller than the small patch-to-patch differences in those thousand-numbered targets?

First of all, thanks for the methodical test. It's very useful.

As for the how many are enough, there are obviously diminishing returns involved. How about the step size differences on an un-scrambled target? I assume that when the stepsize starts coming closer to the natural deviation for the device (on the paper tested!), the only benefit of a larger number of patches is the averaging/canceling out of variations. Whether that justifies the extra paper and time needed remains to be seen.

Cheers,
Bart

[Ernst Dinkla]

Discussions on large numbers of patches in this forum are about unique patches and so far little about repeating patch colors, so when one or more patch readings are compromised it is affecting the profile directly.

Some days ago I hit on a discussion on the optimal patch size for an Eye 1:
http://www.colorforums.com/viewtopic.php?p=4228&sid=d1944efc17591c866eed240a86445938


met vriendelijke groeten, Ernst

New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[Bart_van_der_Wolf]

Discussions on large numbers of patches in this forum are about unique patches and so far little about repeating patch colors, so when one or more patch readings are compromised it is affecting the profile directly.

Assuming you are reacting to my post, I'm not suggesting multiple readings of identical patches by using a large number of patches. The patches will be different, as can be seen on the non-scrambled targets. The issue is, when a patch is read, it will almost certainly not produce an exact reading. In fact it will produce a different reading each time the patch is read, albeit with a small tolerance. When the small tolerance exceeds the step difference, then a sort of averaging takes place of those steps, and the real accuracy of the readings will not produce much better profiles than when fewer patches are used. In fact, the step differences can be quite a bit larger than that, because there will be interpolation (and maybe extrapolation) used to build the TRCs anyway.

As for the scanning travel speed, to detect where one patch starts and another ends, each patch is sampled multiple times during the travel distance. These can also be considerd as multiple readings, but in this case of identical patches. The same logic that determines the boundaries between patches, can use all the samples between the boundaries to average into a statistically more relevant reading. This was demonstrated by the OP, slower travel speed equals more readings per patch which equals higher consistency. Higher hardware speed is a benefit.

The specifications of the i1 on the Xrite pages state a:

Short-term repeatability: DE*94 <= 0.1 (D50,2°), with respect to the mean CIELab value of 10 measurements every 3 seconds on white

which suggests a sampling interval of at most 0.3 seconds, but I assume it's faster because it can differenciate and read more patches than that with high repeatability.

Cheers,
Bart

[Ernst Dinkla]

The 'step difference' term confused me. 

The algorithm to detect the boundaries and get cleaner patch readings could be adjusted for speed, reducing usable patch readings though. I guess it will be hard to detect boundaries, keep usable patch readings and throw out the most extreme readings in the last and then average.


met vriendelijke groeten, Ernst


New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

[aaronchan]

I have been looking into that too. The Z3200 Color Center target import and measurement output is however quite unpredictable, getting it to work is one thing, saving the measured data on the computer another thing. With APS it goes better but that one only exports spectral data. BTW. the spectrometer doesn't make 380, 390 and 410 Nm spectral readings, 400 Nm is included and at the red end it skips the 720 + 730 Nm wavelengths. Is that similar to the UV cut iSis?

met vriendelijke groeten, Ernst

New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm

So far I'm getting very accurate and consistent result with the Color Utility on Mac.

[Ernst Dinkla]

So far I'm getting very accurate and consistent result with the Color Utility on Mac.

Aaron,

It is possible we are not discussing the same software feature. There is no problem with Color Center in its usual target printing, measuring, profile building and saving that profile to system and printer. On both the Z3200 and Z3100. The same with APS. There is another route to use Color Center on the Z3200 for profiling other printers, custom target input, target Tiff export, measuring, data export: it is called Color Measurement with an i1 logo next to it.

When it works the measurements are fine (within the limitations of the spectrometer). The network connection can be a problem though. I have opened up the ports that could block that particular communication. I check the connection first with another Printer Utility feature to be sure.  Meanwhile I can use it without issues. Including the use of greyscale target descriptions for QTR linearisation + profiling. A command line tool doing the same for the Z3100 is however less stable. More picky on the custom target description and network connection. But I get there too.
APS offers a similar feature for the Z3100 and Z3200 but it only exports spectral data which need conversion work to be usable for the QTR linerisation tool.


met vriendelijke groeten, Ernst

New: Spectral plots of +250 inkjet papers:

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm