Pages: [1] 2 3   Go Down

Author Topic: Using the HP Z3200 spectrophotometer to generate profiles for other printers  (Read 19834 times)

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

Hello,

I thought about creating this tutorial several times, but never managed to actually do it. A few days ago Mark Lindquist asked me to post this step-by-step just as I was preparing to make another batch of profiles. As the timing was perfect I decided to do it this time. Please feel free to ask anything, specially if something sounds weird (english is not my native language, so...)

Before we start, let's make some things clear:
  • If I had an x-rite i1io robotic arm to automate the scanning I would never go this route, but it is an expensive piece of hardware even if you already have the i1 spectro. I cannot justify this purchase as I already have an automated i1 spectrophotometer inside my Z3200.
  • Some people think the profiles produced this way are somehow inferior to the profiles produced with a standalone i1 spectrophotometer. I respectfully disagree. The problem may lie on HP software (APS) that most people tend to use to generate the profiles and my solution is to use another software to process the data and to generate the profile. Doing this I could not identify any significant differences between the "automated" profile and the one created with the standalone spectro.
  • One downside is the big patch size required by the Z3xxx system. It is quite large, so you will use a considerable amount of paper to make a 1728 patch chart: on a 24" roll it will use a 24x26" piece of paper [see image 1 attached] or two A2 sheets or nine A4 (or 8.5x11") sheets.
  • Another downside is the amount of work. The only thing automated is the chart reading, the rest requires a lot of user intervention. Don't get me wrong, I still believe it is better than hand scanning the charts once you have digested the process and created your own workflow and shortcuts.
  • This tutorial is based on a system running windows, but the steps and softwares are the same for mac.

HP Z printers may use their HP Utility "color center" or HP's APS software to create profiles. We will use the basic "Color Center" to generate the chart file and to read the printed chart, but another piece of software is needed to create the profile. I like to use Argyll CMS for this, a free color management system, but you may be able to process the TXT file containing the data using any other profile creation software that accepts external files.

The process:
  • Use the  HP Utility's "color center" to generate the TIFF chart files.
  • Print the chart using the desired printer with the desired paper.
  • Read the chart using the HP Z printer and HP Utility's "color center" (it will generate a TXT data file).
  • Process the file containing the data on a profiling software to generate the profile.

Step One: generate the target
Open the "HP Utility" software, go to the "Color Center" tab and select "Paper Preset Management".
Now, instead of selecting the "Profile Paper" option as you usually would to create profiles, you want to select the last option on the right side: "Color Measurement". [see image 2 attached]
Select the last option, "Export a color chart as a TIFF file for external printing", than browse to the desired folder and give the file a name. Click next.
Select the type of chart and number of patches. You want RGB for sure and I usually go for the largest, 1728 patches.
Now you have to specify the paper format (roll or sheet) and the size (width for rolls and width and length for sheets).
The resulting file (or files in case of smaller sheets) will exactly match the paper size specified and will include enough white borders to avoid cropping on most printers.   
Generate files for the sizes you use the most and you will never have to repeat this step. In the future just print the file again on the desired paper and printer.

Step Two: print the chart
No mystery here, just print it as you normally would for a profile chart with the desired printer and paper (no color management, basically). Just make sure to don't enlarge or shrink the chart nor crop the patches in any way. The whole "automated" part depends on this. The TIFF file already includes enough borders to avoid cropping the patches on most printers, just match the image size with the paper size and let the white borders be cropped by the minimum margins of your printer.
Usually I let the targets dry overnight before reading it.

Step three: Read the chart on the HP Z
As long as you have printed the chart correctly, this step should be easy. You just have to load the printed chart correctly:
  • Load the printed chart before starting the reading process.
  • Load it as if it was a blank paper that you want to print on.
  • Insert the top side first, the one with the line of colored rectangles.
  • Select "Sheet, with skew check" when the printer asks.

Tip 1: Load the printed chart through the roll paper feed, not through the cut sheet tray. Have an empty roll core on the axle to use it as a support and guide when feeding large pieces. [see image 3 attached]

(Continues below)
« Last Edit: December 07, 2016, 03:44:16 pm by Geraldo Garcia »
Logged

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

Tip 2: When printing on canvas some printers feed the material a bit more before printing, generating a top margin bigger than what is expected [see image 4 attached]. It may cause reading problems but you may solve this cutting the extra canvas manually (leave the original top margin) or loading it as it is and advancing the media a few clicks before initiating the reading process (stop when you start to see the colored rectangles under the rollers)[see image 5 attached].

With the chart properly loaded, open the "HP Utility" software, go to the "Color Center" tab and select "Paper Preset Management". Once more select the last option on the right side: "Color Measurement". This time select "Measure a color chart previously printed".
Select the type and patch size of the color chart you printed (RBG 1728 patches in my case) and click next.
Now you must choose the name, location and type of the resulting data file. My advice is to give it exactly the name you want for the final profile to keep things simple. I usually go with something like this: "MyInitials_Printer_Paper_date" and for this tutorial use a .TXT extension.
On "File Format" select "CGATS measurement file" and mark all the checkboxes. Leave the band selection on the default 10nm. Click next.

Now the printer will check the position of the chart. If something is wrong it will eject it and abort. If the chart was correctly loaded it will move to the "preparing" phase to warm up the spectrophotometer for a few minutes. After that it will star reading and the whole process may take half an hour or so. If your char is divided on various cut sheets you will need to load the subsequent sheets on the correct order.

After that you will find a .TXT file with all the data saved on the specified location.

Step Four: generate the profile using a profiling software
You are free to use any software you want, but I will exemplify using Argyll CMS, an excellent and powerful software that is free, available for Mac and PC. The bad news is: it is a command-line software, no graphic user interface. But it is not that hard to use once you know the commands.

Google, download and install Argyll CMS.
Copy the .TXT file we created to the "bin" folder of the Argyll CMS.
Now we have to convert the .TXT file to a .TI3 file. Open a command prompt, go to the "bin" folder and type (Note that "FILENAME" is the name you gave to the file without the .TXT extension):
txt2ti3.exe -v FILENAME.txt FILENAME

That will create a file with the same name and a .TI3 extension.

Before we create the profile I must explain something: Argyll profiles can be excellent if you use them properly. They work perfectly with Relative Colorimetric but if you want perceptual there is a catch: You must specify the source colorspace when creating the profile. That means you will have a profile that is usable for conversions from any colorspace using relative colorimetric, but only from one colorspace when using perceptual. If you use perceptual converting from a different colorspace the result will not be as precise as it could be.

That is not a big problem if you standardize your printing workflow, always exporting files using, let's say, ProPhotoRGB before printing. You may also create different profiles to convert from different workspaces (and please name them in a way that makes sense).

So, before the final step you may want to copy the source colorspace file to the "bin" folder. Let's assume we copied the "ProPhoto.icm" to the "bin" folder.

Now it is time to create the profile. On the command prompt type:
colprof.exe -v -qh -S ProPhoto.icm -cmt -dpp FILENAME

It may take a few minutes to process depending on the processing power of your computer. In the end you will have the FILENAME.icm file ready to install and use.

Some final tips:
  • You may want to create shortcuts on your desktop with the command line entries for "txt2ti3" and "colprof". That way you just update de FILENAME when you want to create a new profile and execute the shortcut, avoiding the need to type the commands every time.
  • When you want to create different profiles from the same chart, to use a different source profile for perceptual rendering, just rename the .TI3 file and run the "colprof" again with the desired changes.

That is it! I swear it is not as complicated as it sounds.

Best regards.
« Last Edit: December 07, 2016, 04:37:54 pm by Geraldo Garcia »
Logged

Mark Lindquist

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1596
  • it’s not about the photos we take - it’s the ones we leave
    • LINDQUIST STUDIOS

Fantastic Geraldo!
Great tutorial - very much appreciated by us all.

-Mark
Logged
Mark Lindquist
http://z3200.com, http://MarkLindquistPhotography.com
Lindquist Studios.com

Mark Lindquist

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1596
  • it’s not about the photos we take - it’s the ones we leave
    • LINDQUIST STUDIOS

Geraldo, out of curiosity, do you know a software for PC and Mac that converts the .TXT file automatically without having to use the command line? 

Also, the next to largest size target, being used to make the profile, is there a very big difference between it and the 1728 patches target?

Thanks again,

Mark
Logged
Mark Lindquist
http://z3200.com, http://MarkLindquistPhotography.com
Lindquist Studios.com

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

Geraldo, out of curiosity, do you know a software for PC and Mac that converts the .TXT file automatically without having to use the command line? 

I believe X-Rite's profilemaker can do it, but I have to check it as I don't have it installed on this computer.
Using Argyll CMS the only way to avoid the command line is to create a shortcut icon with that command and parameters and update the filename on the "properties" before using it. You may do it for the file conversion and also for the profile creation.


Quote
Also, the next to largest size target, being used to make the profile, is there a very big difference between it and the 1728 patches target?

The options are 343, 729 or 1728 patches. Every time I thought about making a 729 patch profile I changed my mind before printing, so I can't compare them, actually. I always think: "Since the printer will scan the chart... Hell, let's go for the bigger" ;D
« Last Edit: December 07, 2016, 04:39:46 pm by Geraldo Garcia »
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498

Hi Geraldo,

Do you know what is the repeatability of letting the Z3200 measure the chart? If you have 2-3 measurements of the same chart in CGATs format it would be great if you don't mind sharing. I would be curious to do a dE comparison in ColorThink. One good thing to know is whether it beats the repeatability of measuring by hand (probably). I'm also curious to know how close it comes to an iSis.

Interesting about the 1728 patch limit. I wonder if it is possible to hack the software (maybe by changing one of the source files or system files) to allow for any limit. In theory there should be no limitation to patch size, any arbitrary number could be measured, and custom chart designs could be used.

The limitation you pointed out about Argyll Perceptual gamut mapping method is a critical one. I've been working to encourage Graeme to provide another solution to that for quite some time, but it is not really ready yet. The last two versions of Argyll has already started to offer some version of it, but it really requires several more user-controllable options to make it work right. As it is, it is unusable. I'm in the midst of preparing materials for Graeme. Hopefully we can all have the greatest color gamut mapping the world has never seen soon. ArgyllCMS has many, many good things going for it that make it much better than many other options out there.
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4388
    • Pieter Kers

Many thanks Geraldo,
I will have a look into it...and it starts me off in Argyll- that always looked too complicated to me

I have a Z3100 and wonder if the APS in question is the same or different from the Z3200...

greetings,

PK

Logged
Pieter Kers
www.beeld.nu/la

Damir

  • Full Member
  • ***
  • Offline Offline
  • Posts: 237

It is possible to use more patches, I calibrate with APS on Z3100 with 4096 patches using Bill Atkinson target. He also have target with 5202 patches.

It is all connected to .txt file in APS software which will generate as many patches as you put in it. As I recall all that information you can find here on forum, or Luminous Landscape main page.
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4388
    • Pieter Kers

It is possible to use more patches, I calibrate with APS on Z3100 with 4096 patches using Bill Atkinson target. He also have target with 5202 patches.

It is all connected to .txt file in APS software which will generate as many patches as you put in it. As I recall all that information you can find here on forum, or Luminous Landscape main page.

Are more patches making your profile better?
I have used 1728 patches in the past, considering that a lot already.
I suppose it how to deal with the software writing the profile... i always used the APS for calculating the profile; maybe Argyll has a better way of calculating profiles?



Logged
Pieter Kers
www.beeld.nu/la

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

Do you know what is the repeatability of letting the Z3200 measure the chart? If you have 2-3 measurements of the same chart in CGATs format it would be great if you don't mind sharing. I would be curious to do a dE comparison in ColorThink. One good thing to know is whether it beats the repeatability of measuring by hand (probably). I'm also curious to know how close it comes to an iSis.

I will run this test as soon as I can, probably during the weekend, and will report back..

Quote
The limitation you pointed out about Argyll Perceptual gamut mapping method is a critical one. I've been working to encourage Graeme to provide another solution to that for quite some time, but it is not really ready yet. The last two versions of Argyll has already started to offer some version of it, but it really requires several more user-controllable options to make it work right. As it is, it is unusable.

That depends on your workflow. I use relative colorimetric 90% of the time. When I want to test/use perceptual I just convert the image to prophoto (the source colorspace I used when creating the profiles) before doing it. Actually, most images I print are in prophoto already, so it is not a big deal to me.
Logged

Damir

  • Full Member
  • ***
  • Offline Offline
  • Posts: 237

Are more patches making your profile better?
I have used 1728 patches in the past, considering that a lot already.
I suppose it how to deal with the software writing the profile... i always used the APS for calculating the profile; maybe Argyll has a better way of calculating profiles?

Difficult to say it seems to me that there is some minor improvement, but you know it may be just a placebo :-)
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498

I will run this test as soon as I can, probably during the weekend, and will report back..

Thank you!  :)

Quote
That depends on your workflow. I use relative colorimetric 90% of the time. When I want to test/use perceptual I just convert the image to prophoto (the source colorspace I used when creating the profiles) before doing it. Actually, most images I print are in prophoto already, so it is not a big deal to me.

It's not about workflow, it's about gamut mapping. The gamut mapping of OOG colors for Rel Col rendering is not user-adjustable, and the way it is done by Argyll and every other solution is to do "colorimetric minimized absolute color error" - the OOG color is mapped to the closest possible color. While it sounds unintuitive, doing this usually hurts regular photographs. Most of the time preserving lightness over chroma causes OOG to shift less visually.
Logged

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

It's not about workflow, it's about gamut mapping. The gamut mapping of OOG colors for Rel Col rendering is not user-adjustable, and the way it is done by Argyll and every other solution is to do "colorimetric minimized absolute color error" - the OOG color is mapped to the closest possible color. While it sounds unintuitive, doing this usually hurts regular photographs. Most of the time preserving lightness over chroma causes OOG to shift less visually.

That is the whole debate about Colorimetric versus perceptual, but it may be a matter of workflow. If you print images that are inside the gamut or 98% inside the gamut and the other 2% are not relevant to the image, Relative Colorimetric is perfect. And don't forget: if you want to use perceptual with Argill profiles, you can. You just have to pay attention to do source colorspace.
Logged

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

Do you know what is the repeatability of letting the Z3200 measure the chart? If you have 2-3 measurements of the same chart in CGATs format it would be great if you don't mind sharing. I would be curious to do a dE comparison in ColorThink. One good thing to know is whether it beats the repeatability of measuring by hand (probably). I'm also curious to know how close it comes to an iSis.

Hello Samuel,

I measured the same chart 3 times with the Z3200 allowing an interval of a couple of hours between the readings. Here is the DeltaE2000 report comparing the readings:
Quote
--------------------------------------------------
dE Report

Number of Samples: 1728

Delta-E Formula dE2000

Overall - (1728 colors)
--------------------------------------------------
Average dE:   0,09
    Max dE:   0,51
    Min dE:   0,00
 StdDev dE:   0,06

Best 90% - (1554 colors)
--------------------------------------------------
Average dE:   0,07
    Max dE:   0,16
    Min dE:   0,00
 StdDev dE:   0,04

Worst 10% - (174 colors)
--------------------------------------------------
Average dE:   0,23
    Max dE:   0,51
    Min dE:   0,16
 StdDev dE:   0,07
--------------------------------------------------

I was not expecting such excellent repeatability. I know it is better than what we get hand scanning with de i1 pro II, but I don't know about the iSis.
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498

Hello Samuel,

I measured the same chart 3 times with the Z3200 allowing an interval of a couple of hours between the readings. Here is the DeltaE2000 report comparing the readings:
I was not expecting such excellent repeatability. I know it is better than what we get hand scanning with de i1 pro II, but I don't know about the iSis.

Hi Geraldo,

Thank you for the information. I assume that the first reading pass was after sufficient dry down time, not right after the chart was printed? Then the several hours between readings should not matter enough. It would be good to know that the dE differences is almost entirely due to the instrument's limitation, and we are not looking at dry down changing the dE also.

Otherwise the repeatability is about the same as what one can get with careful hand scanning with the i1 Pro II, in my tests. The repeatability is limited by the measuring device in this case - the same i1 Pro base technology. However it is less likely for the automated Z3200 system to make errors due to hand slippage or inconsistent or too fast/slow movement, so human error is minimized. Occasionally I catch an error of about 1 dE2000 when hand scanning in the past. Sometimes a freak accident is much worse. It was good to see confirmation from Damir that it is possible to hack the software to scan any number of patches. If I had a Z3200 and no iSis I would use it for scanning too.

The iSis repeatability is on a whole different level. Here is a comparison of scanning a (more challenging) 2808 patch target. The max dE difference is more than an order of magnitude better and the same could almost be said of the average dE:

Quote
Number of Samples: 2808

Delta-E Formula dE2000

Overall - (2808 colors)
--------------------------------------------------
Average dE:   0.01
    Max dE:   0.03
    Min dE:   0.00
 StdDev dE:   0.00

Best 90% - (2526 colors)
--------------------------------------------------
Average dE:   0.01
    Max dE:   0.01
    Min dE:   0.00
 StdDev dE:   0.00

Worst 10% - (282 colors)
--------------------------------------------------
Average dE:   0.02
    Max dE:   0.03
    Min dE:   0.01
 StdDev dE:   0.00


Quote
That is the whole debate about Colorimetric versus perceptual, but it may be a matter of workflow. If you print images that are inside the gamut or 98% inside the gamut and the other 2% are not relevant to the image, Relative Colorimetric is perfect.

There is no debate here, and I am not trying to incite one. It's not about if Colorimetric or Perceptual is better, it is about how our visual system processes colors and what changes are least obvious when performing gamut mapping. Gamut mapping should not be an aesthetic tool for creative adjustment, it should only strive to minimize visual differences. Neither Colorimetric nor Perceptual rendering is good enough in all the profiles generated by today's commercial software. Here's something interesting for pondering: No one ever said how OOG colors should be treated in Colorimetric rendering. We can look around and see different software packages rendering OOG colors in Rel Col differently. It was possible in a very old software, Color Savvy, to create Perceptual mapping that is exactly like Colorimetric for the in-gamut colors but the OOG color treatment was better, and there were controls to tweak the OOG rendering in beautiful ways. You must have noticed that when soft proofing on matte media, switching between the Rel Col and Perceptual rendering intent the preview changes a lot (have paper white and black ink unchecked). Also how different the tonality is for Perceptual rendering between glossy and matte papers, but not Rel Col. Why should it be so different? It is not in the best examples of Savvy's profiles. I would be just as happy to use Rel Col if BPC worked perfectly all the time and worked better (cause less hue shifting), and if we had options to dictate how OOG colors are mapped, I could care less about Perceptual. We really need more options for controlling OOG color mapping when creating printer profiles.

Quote
And don't forget: if you want to use perceptual with Argill profiles, you can. You just have to pay attention to do source colorspace.

Not if one is using the new "-lp" luminance-preserving mapping intent for generating the Perceptual table, which I suggested to Graeme to offer. We have spent a lot of time in discussion about these issues and working towards getting this right. A word of caution, it is not yet working properly so I don't advise to use it. The tone curve issues are resolved and it is now possible to have a tone curve that is even somewhat better than Rel Col - they are very close except for a tiny difference in the shadow tonality roll-off. The available controls which were not designed for this job happen to work brilliantly regardless of the dmax of the media. The in-gamut color mapping is resolved. There are still some ways to go before the mapping is truly luminance-preserving and we also need a control that then lets us trade off OOG colors lightness vs chroma preservation, so a bouquet of profile variants for each printer+paper combo can be made for a variety of OOG color mapping needs.
« Last Edit: December 12, 2016, 09:33:08 pm by samueljohnchia »
Logged

Ernst Dinkla

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4005


Otherwise the repeatability is about the same as what one can get with careful hand scanning with the i1 Pro II, in my tests. The repeatability is limited by the measuring device in this case - the same i1 Pro base technology. However it is less likely for the automated Z3200 system to make errors due to hand slippage or inconsistent or too fast/slow movement, so human error is minimized. Occasionally I catch an error of about 1 dE2000 when hand scanning in the past. Sometimes a freak accident is much worse. It was good to see confirmation from Damir that it is possible to hack the software to scan any number of patches. If I had a Z3200 and no iSis I would use it for scanning too.

The iSis repeatability is on a whole different level. Here is a comparison of scanning a (more challenging) 2808 patch target. The max dE difference is more than an order of magnitude better and the same could almost be said of the average dE:


There is a reason why the patches are that big on the normal HP Z targets. The spectrometer is not as close to the patches as the Isis etc instruments are. Repeated measurements of the same target (more times loaded on the printer) do not deliver exactly the same spots measured within the patches. First because there is some tolerance on the register of the target sheet in the loading phase, second because the sheet might expand or contract between measurements. With smaller targets, smaller patches and higher precision of the sheet loading this fault could diminish.

On the other hand I was no hero in manual measurements of big targets so I'm happy the repeatability is equal to careful manual scanning with the i1 Pro II.

Averaging several measurement results before creating the profile should be possible with APS.


Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
November 2016 update, 700+ inkjet media white spectral plots
Logged

samueljohnchia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 498

There is a reason why the patches are that big on the normal HP Z targets. The spectrometer is not as close to the patches as the Isis etc instruments are. Repeated measurements of the same target (more times loaded on the printer) do not deliver exactly the same spots measured within the patches. First because there is some tolerance on the register of the target sheet in the loading phase, second because the sheet might expand or contract between measurements. With smaller targets, smaller patches and higher precision of the sheet loading this fault could diminish.

On the other hand I was no hero in manual measurements of big targets so I'm happy the repeatability is equal to careful manual scanning with the i1 Pro II.

Averaging several measurement results before creating the profile should be possible with APS.

That is more or less similar to what happens when using the Canon SU-21 spectro accessory for iPF64X0 printers. Do you know the physical size of the patches in mm, edge to edge of the hexagon, not tip to tip?

The iSis has incredible repeatably not because of the closer positioning height of the measurement aperture - in fact the i1 Pro sits closer to the sheet than the measuring head in the iSis. The main advantage over the i1 Pro is how it detects patches. The iSis uses the same i1 spectro technology and indeed in spot read mode the i1 Pro is as repeatable as the iSis. The i1 Pro in scan mode, makes 200 measurements per second and uses an algorithm which throws out bogus measurements made when the measurement aperture is over two adjacent patches. This detection is obviously not ideal for several reasons, but it is simple and works well enough for most. On the iSis, patch detection is physically done in both horizontal and vertical direction. The iSis knows the boundary of where each patch is because it knows what size the patches are (input in i1Profiler), it knows how far it advances the target and it also knows where the measurement head carriage is positioned. There is also some compensation for skew errors, but it is not foolproof and must be avoided as far as possible. As a result, all corrupted patch measurements are thrown out totally reliably.

I don't know which method the on-board spectro on the Z3200 uses for patch detection. If it does not have horizontal physical positioning detection, the repeatability would be mainly limited by software detection of patches, assuming no paper feeding issues. It would be very hard to get it below the already excellent performance that Geraldo has shown.

It should be safe to assume that the distance to the measurement aperture remains more or less the same for stiff papers, over a relatively short period of time, and also after the paper as been allowed to acclimatize to the room conditions. The pinch rollers and vacuum should keep the paper nice and flat over the platten for measurement? I'm not familiar with the Z3200. Paper warping does affect the iSis too and should be avoided as far as possible. The precision of sheet loading is not perfect, which is why I use very large patch sizes for my custom iSis targets. Smaller patches increases the risk of measurement errors, not reduces them.
Logged

Ernst Dinkla

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 4005

That is more or less similar to what happens when using the Canon SU-21 spectro accessory for iPF64X0 printers. Do you know the physical size of the patches in mm, edge to edge of the hexagon, not tip to tip?

The iSis has incredible repeatably not because of the closer positioning height of the measurement aperture - in fact the i1 Pro sits closer to the sheet than the measuring head in the iSis. The main advantage over the i1 Pro is how it detects patches. The iSis uses the same i1 spectro technology and indeed in spot read mode the i1 Pro is as repeatable as the iSis. The i1 Pro in scan mode, makes 200 measurements per second and uses an algorithm which throws out bogus measurements made when the measurement aperture is over two adjacent patches. This detection is obviously not ideal for several reasons, but it is simple and works well enough for most. On the iSis, patch detection is physically done in both horizontal and vertical direction. The iSis knows the boundary of where each patch is because it knows what size the patches are (input in i1Profiler), it knows how far it advances the target and it also knows where the measurement head carriage is positioned. There is also some compensation for skew errors, but it is not foolproof and must be avoided as far as possible. As a result, all corrupted patch measurements are thrown out totally reliably.

I don't know which method the on-board spectro on the Z3200 uses for patch detection. If it does not have horizontal physical positioning detection, the repeatability would be mainly limited by software detection of patches, assuming no paper feeding issues. It would be very hard to get it below the already excellent performance that Geraldo has shown.

It should be safe to assume that the distance to the measurement aperture remains more or less the same for stiff papers, over a relatively short period of time, and also after the paper as been allowed to acclimatize to the room conditions. The pinch rollers and vacuum should keep the paper nice and flat over the platten for measurement? I'm not familiar with the Z3200. Paper warping does affect the iSis too and should be avoided as far as possible. The precision of sheet loading is not perfect, which is why I use very large patch sizes for my custom iSis targets. Smaller patches increases the risk of measurement errors, not reduces them.

The hexagons measure 15mm side to side, with deviations of 0.2mm though. Anyway 10 patches give 150mm in the three symmetry axis. There is a square patches pattern possible too with again 15mm from side to side.

Observing the movement I assume the positioning is numerical, so physical and not based on throwing out odd numbers when it hits a patch boundary. Based on that and possible paper warping or small deviations in the patch's density I think it could be possible that there are differences between measurements, despite the wider area it measures. Which begs the question whether it measures 600 PPI or 300 PPI printed patches. I assume the last or HP thought 450 PPI would be a nice value in between.  I do not expect it ever measures over a patch boundary.

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
November 2016 update, 700+ inkjet media white spectral plots

Logged

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

I assume that the first reading pass was after sufficient dry down time, not right after the chart was printed? Then the several hours between readings should not matter enough. It would be good to know that the dE differences is almost entirely due to the instrument's limitation, and we are not looking at dry down changing the dE also.

Of course! Or should I say "yes, unfortunately."? I mean, I would love to improve even more the Delta E but the chart dried for two days before reading, so I don't think the wait between the readings made any difference.
Actually there is no need of "hacking" to use different charts, the color center gives the option to "add" the data needed to create a new test chart, I just never tried and forgot to mention it.

The iSis is really on another league in terms of repeatability! That is truly impressive and made me consider it (once again) as an alternative for the day my Z3200 gets beyond repair. As you use it regularly please tell me how it behaves with thick media, like rough aquarelle paper of 310~350gsm? I heard some conflicting opinions about it in the past.

Quote
(...)I would be just as happy to use Rel Col if BPC worked perfectly all the time and worked better (cause less hue shifting), and if we had options to dictate how OOG colors are mapped, I could care less about Perceptual. We really need more options for controlling OOG color mapping when creating printer profiles.

I agree 100%.

Quote
Not if one is using the new "-lp" luminance-preserving mapping intent for generating the Perceptual table, which I suggested to Graeme to offer. We have spent a lot of time in discussion about these issues and working towards getting this right. A word of caution, it is not yet working properly so I don't advise to use it. The tone curve issues are resolved and it is now possible to have a tone curve that is even somewhat better than Rel Col - they are very close except for a tiny difference in the shadow tonality roll-off. The available controls which were not designed for this job happen to work brilliantly regardless of the dmax of the media. The in-gamut color mapping is resolved. There are still some ways to go before the mapping is truly luminance-preserving and we also need a control that then lets us trade off OOG colors lightness vs chroma preservation, so a bouquet of profile variants for each printer+paper combo can be made for a variety of OOG color mapping needs.

That is something I am anxious to see working properly. Will give it a try just to see what happens.

Quote
I don't know which method the on-board spectro on the Z3200 uses for patch detection. If it does not have horizontal physical positioning detection, the repeatability would be mainly limited by software detection of patches, assuming no paper feeding issues. It would be very hard to get it below the already excellent performance that Geraldo has shown.

I don't know how it handles it internally, but I can tell you what I see: The spectrophotometer does not move in a continuous motion, it moves and stops over each patch to read it. There are also "code bars" at the beginning of each line of hexagons and I believe it helps the system to align properly.

You asked about the size of the hexagons: 15mm would be the maximum diameter of a circle contained inside of the hexagon.

Regards.
Logged

Geraldo Garcia

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 470
    • Personal blog

(...)Which begs the question whether it measures 600 PPI or 300 PPI printed patches. I assume the last or HP thought 450 PPI would be a nice value in between.  I do not expect it ever measures over a patch boundary.

Hello Ernst,

Checking the TIFF file color center generates for us to print, it has 300 PPI on the final print size, so I guess 300 PPI is what the system uses.

Regards.
Logged
Pages: [1] 2 3   Go Up