Pages: [1] 2   Go Down

Author Topic: Characterizing the P20000 and creating an optimizing patch set  (Read 4177 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197

This is a joint effort between narikin and I to:

1. Characterize the Epson P20000 printer
2. Then create a patch set to optimize profile accuracy with a focus on dark colors.

The patch set has 2871 RGB values containing:
1. 1728 patches on the grid with values 0, 10, 25, 40, 70, 100, 130, 160, 190, 220, 245, 255
2. All neutrals and near neutrals on a 37 point grid to correspond to the I1P profile structure
3. Repeated colors (6x) over the 27 RGB colors spaced 85 apart not including 0's
4. All neutral colors spaced 3 apart RGB 0/0/0, 3/3/3, 6/6/6 .... 255/255/255
5. Repeated deep shadow colors until total patch set equals 2871

The primary purpose of this patch set is to gather information on the printer driver's color mapping. Printing and scanning these patches provides statistical data on the printer's color consistency when printing the same colors in varying locations. It also provides much information on the smoothness of the neutrals tone curve. These provide the info required to create a patch set that will generate a profile specific to that printer/paper type.

Narikin, who has an iSiS and P20000 has kindly offered to engage in this experiment. For me, I know the day will come when my trusty 9800 will go belly up and this gives me insight into whether I may upgrade to the P10000, the 44" version otherwise compatible.
Logged

Royce Howland

  • Full Member
  • ***
  • Offline Offline
  • Posts: 155
    • Vivid Aspect Photography
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #1 on: April 19, 2018, 11:08:12 am »

Guys, if another set of data would be of interest, I'd be happy to join in this exercise. We have a P10000 and a P20000 that could be run through the same profiling workflow, albeit on the scanning side with an i1 Pro 2 and associated scanning table rather than an iSis.

I was already planning to invest some effort in coming weeks, to see whether we can optimize our profiling workflow; not only with the big SureColors but also with our new Canon PRO-6000. I wouldn't say I'm unhappy with the output of our current i1Profiler results (using the standard 2371 patch layout), but neither am I fully satisfied that it's delivering everything that these new inksets are capable of. I'm looking to satisfy myself on whether we're reasonably close to peak performance in terms of gamut, Dmax, linearity of neutrals and good separation of dark colours and greys. (Recognizing of course that there may be some trade-offs in these factors.)

If you prefer to keep focus on a single machine for now, that's cool. I'll carry on with my own efforts regardless, and be very interested to see what transpires with yours.

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #2 on: April 19, 2018, 11:36:22 am »

Guys, if another set of data would be of interest, I'd be happy to join in this exercise. We have a P10000 and a P20000 that could be run through the same profiling workflow, albeit on the scanning side with an i1 Pro 2 and associated scanning table rather than an iSis.

I was already planning to invest some effort in coming weeks, to see whether we can optimize our profiling workflow; not only with the big SureColors but also with our new Canon PRO-6000. I wouldn't say I'm unhappy with the output of our current i1Profiler results (using the standard 2371 patch layout), but neither am I fully satisfied that it's delivering everything that these new inksets are capable of. I'm looking to satisfy myself on whether we're reasonably close to peak performance in terms of gamut, Dmax, linearity of neutrals and good separation of dark colours and greys. (Recognizing of course that there may be some trade-offs in these factors.)

If you prefer to keep focus on a single machine for now, that's cool. I'll carry on with my own efforts regardless, and be very interested to see what transpires with yours.

Sure, the more the merrier (within reason). I have program scripts that automatically analyze the scan results, While I expect variations due to instrument differences and, of course there are slight media differences within type, it would be interesting to see if the statistics are similar. One thing I request, is that all measurements by in M2 mode. This is the default iSiS mode. However, what comes out of this should be quite good for any mode.

Just save the measurement data in the I1Profiler's default format and include it as an attachment here.
Logged

Royce Howland

  • Full Member
  • ***
  • Offline Offline
  • Posts: 155
    • Vivid Aspect Photography
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #3 on: April 19, 2018, 11:46:29 am »

I typically use dual scanning mode for all readings and generate profiles out of that using M2. But I can easily scan in M2 for this exercise. What papers are you looking at? And do you want to control for things like pre-scan drying time or other factors?

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #4 on: April 19, 2018, 12:19:17 pm »

I typically use dual scanning mode for all readings and generate profiles out of that using M2. But I can easily scan in M2 for this exercise. What papers are you looking at? And do you want to control for things like pre-scan drying time or other factors?
If you use an I1 Pro 2 you have to do two scans to get M2 profiles. The device is not uV cut and reads M0 (partial uV) by default. It then uses the second pass with a uV LED to read the fluorescent response and synthesizes M1 and M2 from the two readings.

However, if you have an I1 Pro uV cut spectro it will generate M2 profiles natively. Also, if the paper has zero or very low OBAs then an M0 scan is fine since it will not be very different from an M2.

So whichever works to make M2 readings is fine.
« Last Edit: April 19, 2018, 12:27:24 pm by Doug Gray »
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #5 on: April 19, 2018, 01:18:34 pm »

Thanks Doug,

I'll run the patches today, let them dry overnight (at least) and read them soon after.
Will PM you the data in due course.
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #6 on: April 19, 2018, 01:41:21 pm »

Dropping the pxf file into i1 profiler it says patch size is not specified and sets default. Did you use 6 or 7mm patches?

I was thinking this would fit in a single page 'Super B' but it didn't at default settings. I'll play with it a bit.
Edit: got it to fit at default.

Scrambled patches, right?

And if anyone else is doing this, I'm using the latest Epson driver, #18913, as printer driver is a factor in all this.
« Last Edit: April 19, 2018, 01:48:35 pm by narikin »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #7 on: April 19, 2018, 04:24:03 pm »

Dropping the pxf file into i1 profiler it says patch size is not specified and sets default. Did you use 6 or 7mm patches?

I was thinking this would fit in a single page 'Super B' but it didn't at default settings. I'll play with it a bit.
Edit: got it to fit at default.

Scrambled patches, right?

And if anyone else is doing this, I'm using the latest Epson driver, #18913, as printer driver is a factor in all this.

It's set up to print 3, 8.5x11 sheets. But it doesn't matter. Feel free to format it in whatever style you typically use. Yes, it's scrambled. Need to do that for the statistical info. There are a lot of duplicated patches in special areas that I use to collect variance info. I use 6x6mm. More than sufficient for this.
Logged

ZOG

  • Newbie
  • *
  • Offline Offline
  • Posts: 39
    • http://
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #8 on: April 22, 2018, 03:26:03 pm »

Do you print directly from photoshop or via a RIP? I just got mine this past week. It's a lot slower to start printing than my old 11880. Always keeps on cleaning...

I print mostly via Postershop Production House.

I have an i1 Pro2 publish if you want me to make some test on my side. What paper are you using?
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #9 on: April 22, 2018, 11:52:31 pm »

Do you print directly from photoshop or via a RIP? I just got mine this past week. It's a lot slower to start printing than my old 11880. Always keeps on cleaning...

I print mostly via Postershop Production House.

I have an i1 Pro2 publish if you want me to make some test on my side. What paper are you using?

I use the driver direct from Photoshop so the profiles are RGB printer profiles. I've found the 9800, with good profiles, produces very stable, consistent print colors. What I'm really surprised at is how little drift the machine has had in the 10+ years I've owned it.

Feel free to download the target patch set if you wish. It's currently formatted for 3 letter size sheets for the iSis but you can easily just reformat it to I1Pro 2 format. It just is a lot of sheets. But if you want to do the work I can pretty easily run it. Do you also have a new P20000?

After you complete the measurements, hit the "save" button and save the file somewhere convenient. Then just located it at attach it to a post in this thread.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #10 on: April 23, 2018, 12:10:26 am »

I ran this process and made a smaller, but more tailored to the 9800, target consisting of 1914 patches which fit in 2 US letter sheets.

I then made a test target in 16 bit ProPhoto RGB containing 556 patches. These are all the colors in LAB values from L=10 to 100 steps of 10 and a* /b* ranging as far as possible across the gamut, also in steps of 10. These values are uncorrelated with the RGB patches the profile was made with. The patch colors are all those within gamut for both the 9800 and 9500 so I can use the same image for checking both printers.  The ProPhoto target was printed in Photoshop using Abs. Col. then scanned with the iSis.

The dE00 statistics: Ave dE00 .51,  Max dE00 .98

I also printed 4 copies of the neutral Lab values from L=1 to 100 steps of 1. The patches were scanned, duplicates averaged, and the average dE00 was .31 for the in-gamut colors (L=5 to 95). Max dE00 was .8.  The neutral colors are the hardest with the 9800 because dE00 is more sensitive to changes near neutrals and the printer exhibits fairly rapid shifts along the yellow/magenta line. This is a major reason the target has about 400 neutral and near neutral patches.

While I was able to achieve a very slightly better result with 3.8k patches (4 sheets) I am now comfortable that this smaller patch set specific to the 9800 will fill my needs for non-matte prints. It may also work for matte prints but I don't expect to switch the 9800 to MK any time soon.

I've attached a CGATs file for the 1914, 9800 specific RGB target set. It can be loaded into I1Profiler with the  load button and select CGATs file type. It can them be formatted into I1Pro, I1Pro2, or iSis target sheets. Fits nicely on the iSis 957 patches per page US letter size in Profile orientation.
« Last Edit: April 23, 2018, 12:23:10 am by Doug Gray »
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #11 on: April 23, 2018, 08:53:51 am »

Doug,

Here is a link to the iSiS read of your traits file (its too big to attach here). P20000, PK inks, latest driver, max quality, no color management.

Hope this is helpful: https://drive.google.com/open?id=15iQF_Mio6uZNujvnZ3QYlFfDUwoxjtmQ

Thanks again.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #12 on: April 23, 2018, 02:07:41 pm »

Doug,

Here is a link to the iSiS read of your traits file (its too big to attach here). P20000, PK inks, latest driver, max quality, no color management.

Hope this is helpful: https://drive.google.com/open?id=15iQF_Mio6uZNujvnZ3QYlFfDUwoxjtmQ

Thanks again.

Excellent! I downloaded the file and it has what I need. Made a profile from it. General comments, it's somewhat better in the neutrals than the 9800 but, like it, has deviations along the yellow/magenta axis.

General comments:
1. Black point is at L*=2.5
2. Gamut exceeds the 9800 over almost all L*s particularly in the blue->magenta->red->-orange->yellow->green, only slightly less in the cyans between L*20 to L*45. But the differences there are only about 1 dE00.
3. Generally, good looking profile as is with that chart. Average dE00 over the neutral axis: .25. This is not the same as the average dE00 when printing a separate reference file using a profile made from a target file but rather the difference between the profile's report of neutral RGB lab values and the actuals on from the target. It will likely increase somewhat when printing an independent target. That will be the next step after creating an optimized patch set for this printer since this patch set is already heavy on the neutrals and near neutrals.

Next up I'll be looking at the statistics in some depth but impressions are positive.

I presume you normally use 7mm by 7mm patch sizes on 13x19" paper? I'll make a profile optimized for that unless I hear otherwise.

The new one should fit in the 4k limit but if you zip the file it will be much smaller.
« Last Edit: April 23, 2018, 02:34:20 pm by Doug Gray »
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #13 on: April 23, 2018, 02:38:40 pm »

Glad it looks good.

Yes 7x7mm normally, but happy to drop down to 6.5 or 6mm if needed.

Look forward to a profile target as and when, and more importantly, seeing if there is any real difference in the final prints!

Thanks

Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #14 on: April 24, 2018, 05:50:39 pm »

Attached are two Test Chart files formatted for iSis XL, Super B sheets. It consists of 2337 patches, 7mm x 7mm per sheet. The two sheet version is a re-scramble of the first page. I've found this much more effective than re-reading the same targets and averaging because the iSis XL errors are minimal compared to variation in the same RGB patches printed in different locations. Further, just printing the same target multiple times, , averaging with I1Profiler, while somewhat useful, does not do any averaging out of color changes due to being printed at different locations. This latter effect is highly dependent on how well the various settings, platen gap, vacuum, etc. match the paper being used but is always present. The two page chart causes I1Profiler to automatically average the same RGB values even though printed in different physical locations.

The P20000 has some interesting traits. It's neutral axis (R=G=B) has far less variation in b* than my 9800 and is quite consistent from L*=30 and up. However, it does have some variation on the neutral axis at lower RGB values and needs additional neutrals and near neutrals from about R=G=B of 0 to 90. I've added a full set of near neutrals at the I1Profiler high quality grid points plus optimizing patches from the I1Profiler which does a good job filling in RGB values in areas of high 3D interpolation "bending."

This approach did a bang-up job on my 9800 though it required additional patches in different areas. It printed all 500+ reference, in gamut, iSis US letter formatted Lab patches using Abs. Col. from a ProPhoto RGB 16 bit image with less than 1dE00 error and an average dE00 of only .51.

The two charts below show the response of the 9800 and P20000 printers to neutral R=G=B values from 0 to 255. The P20000 has significantly smoother response on both the L* and b* ranges. However, the P20000 does have some variation in a* and b* at lower RGB levels. So I added extra neutrals and near neutrals in this region to these targets.
« Last Edit: April 24, 2018, 06:04:17 pm by Doug Gray »
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #15 on: April 24, 2018, 08:06:27 pm »

Great, Thanks Doug.

I'll look forward to profiling the paper(s), and seeing how this changes things in the prints, if at all.

Very interesting to see the improved response, for the large part, from 9800 to 10000/20000 range. It is not that surprising given almost a decade of progress, and 4 grays in the inkset, but, logic does not always follow through! As you point out it is not a perfect response, and it would be interesting to know if the Canon's are any better in this regard or not - but that's for another thread, another time.

How would those wobbles around the lower end of a* and b* manifest in printed images? - I do see flesh tones of people in shadows having some strange shifts, mostly to red, that need a curve and mask to correct, but that may be the IQ100 response, (which is not a Trichromatic) rather than the profile. We'll see.

Sadly I'm away for the next 6 days, but will report back in a week or so, once profiles are built and applied to a variety of images.
Thanks again for your help with this - are you a color management professional, a math nerd, or simply a very numerate amateur?!





« Last Edit: April 25, 2018, 09:43:02 am by narikin »
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #16 on: April 25, 2018, 11:40:59 am »

Just to confirm I understand this right: the 2 page version is just the 1 page version with the same patches, but in different positions, right?
The idea being to minimize position errors, I believe.

I see a lot more darker tone patches in there, which makes sense with what we are trying to do here.

Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #17 on: April 25, 2018, 01:21:41 pm »

Just to confirm I understand this right: the 2 page version is just the 1 page version with the same patches, but in different positions, right?
The idea being to minimize position errors, I believe.

I see a lot more darker tone patches in there, which makes sense with what we are trying to do here.

All correct.
Logged

narikin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1371
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #18 on: May 04, 2018, 11:32:35 am »

Just to bring this back online.

I managed to output and profile the target. Attached is a sample profile from i1 Profiler:
P20000, Epson Driver set to Legacy Platine, Printer is set to PLU260 (there is no 'Platine' option on printer paper settings - an anomaly) M2 single scan (OBA free) 'Sat' setting (I have Colorful and Neutral iterations)

I do not own Color Think Pro, so can't analyse the profile.
Have not been able to print a big output at full size as yet, so that vital test must wait a while: 'The proof is in the pudding'.
« Last Edit: May 04, 2018, 12:05:18 pm by narikin »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2197
Re: Characterizing the P20000 and creating an optimizing patch set
« Reply #19 on: May 07, 2018, 12:07:56 am »

Just to bring this back online.

I managed to output and profile the target. Attached is a sample profile from i1 Profiler:
P20000, Epson Driver set to Legacy Platine, Printer is set to PLU260 (there is no 'Platine' option on printer paper settings - an anomaly) M2 single scan (OBA free) 'Sat' setting (I have Colorful and Neutral iterations)

I do not own Color Think Pro, so can't analyse the profile.
Have not been able to print a big output at full size as yet, so that vital test must wait a while: 'The proof is in the pudding'.

I downloaded this patch set for creating a profile.I made a profile from it. The profile shape is excellent with nice, non-ragged boundaries.  I also ran the previous data that you posted a few weeks ago. It has a set of RGB values that do not appear in the current profile set that are each repeated 6 times. I ran the average of each of set of 6 through a reverse lookup to get the profile's estimate of Lab values. The average dE76 of the set of averages was quite low, less than .5.  This is a good indication the profile is accurate and that the printer changed little but see below.

However, there is significant variation in Lab values between the 6 patches of identical RGB values. I have seen this as well on my 9800 and it turned out that increasing the paper thickness setting in the driver from .2mm to .3mm significantly reduced this variation. That's one reason I like to randomize patches as well as add extras in critical areas like around the neutral axis.

As for the current patch set, since each patch was duplicated I also compared each corresponding RGB value to its twin. The maximum difference occurred on a saturated red color with a dE76 of 5.0. However, the dE00, because the color was so saturated, was only 1.3.

I also made a low L* (30 down to the black point) patch set of in gamut colors. These turned out to have an ave dE76 of .4 and max of 1.3 on my 9800 which was somewhat better than I expected because low L* colors have a lot of in gamut colors that push way out into the dark blue/violets.


Logged
Pages: [1] 2   Go Up