iSis newbie - how many patches +tips please!

[narikin]

My iSis XL arrived the other week, and now I'm getting around to making my first charts for Epson and Canon printers with i1 profiler.

Anyone recommend a good chart size for me to get high level profiles?
I'm not trying for a quick read, and not scared of multiple pages/ large sizes, but equally no need to go overboard if its pointless

any advice as to where the sweet spot of test patches is? 
tips like whether to randomise them, and/or do dual reads also welcomed!

thanks

[smilem]

With iSis you can do 4000 patches without problems, doing more patches makes little sense.
Use colorport v2 to generate some nice targets, read the targets, save as profilemaker cgats XRGA, load in into i1Profiler and you will have very nice profile.

[shewhorn]

With iSis you can do 4000 patches without problems, doing more patches makes little sense.

I'm not sure if you're saying it doesn't make sense to do 4000 patches or if you're saying, do 4000 patches but more wouldn't make sense?  Fact is, 4000 patches in many cases will result in a terrible profile depending upon what you're profiling. I've found the sweet spot to be around 1200ish patches +/- a few hundred for the current and previous generation of printers.

Use colorport v2 to generate some nice targets, read the targets, save as profilemaker cgats XRGA, load in into i1Profiler and you will have very nice profile.

If you're using i1Profiler I would not recommend generating profiles in Colorport. Why create extra work for yourself and why use something that isn't optimized for use with i1Profiler? If you use the typical patch set values of 729 or 1728 (which magically happen to correspond to 9^3 and 12^3) you probably won't see any different but, you'll notice that in addition to walking RGB calculated numbers, the charts that i1Profiler generates have a number of patches that are close to grey but slightly out. These patches help i1Profiler a great deal in making neutral profiles. In addition to that... why go through all the trouble of reading things in with ColorPort when i1Profiler supports the iSis natively? I might be mistaken here but if you're measuring a paper with OBAs I don't think ColorPort is capable of taking advantage of the dual filtered and unfiltered measurements that i1Profiler uses for OBA/FWC compensation.

Generate your charts in i1Profiler, save them. print them with something else (I still find that i1Profilers printing capabilities can be a bit funky), and read the chart in directly using i1Profiler. The only reason to go to ColorPort (as far as I know) is if you have a device that is not supported (in other words, if you don't have an Eye One Pro or an iSis). I use a Spectroscan/Spectrolino package so ColorPort is the best way to get that data in but even then, the numbers I plug in to ColorPort come directly from i1Profiler.

In terms of finding that sweep spot with regards to the number of patches, I recommend experimenting. There is not magic number.

Cheers, Joe

[narikin]

Fact is, 4000 patches in many cases will result in a terrible profile depending upon what you're profiling. .

Just for my enlightenment, why would 4000 patches give a "terrible profile" over say 1100? (on a modern Epson or Canon LF)

I wouldn't also mind some advice on dual filtered readings with iSis - how this works. I understand about OBA's and UV, filtered/unfiltered etc, but the practicalities of building a OBA-inclusive profile in i1Profiler on the double measurements has been beyond me so far.  For artworks shown in galleries with tungsten lighting, UV is minimal, but many galleries have some daylight, partially activating those OBAs, if present, so where do you draw the line?

[tony22]

If you use the typical patch set values of 729 or 1728 (which magically happen to correspond to 9^3 and 12^3) you probably won't see any different but, you'll notice that in addition to walking RGB calculated numbers, the charts that i1Profiler generates have a number of patches that are close to grey but slightly out.

Joe, I thought in i1Profiler the idea was to use a number that wasn't a multiple of 3?

[shewhorn]

Joe, I thought in i1Profiler the idea was to use a number that wasn't a multiple of 3?

Yes, that's correct. I was just pointing out that the patch generation is similar in multiples of 3 so if you use those patch sets you'll get similar results (and won't get the advantage of the additional patches that help with producing a more neutral grey). A few of the (theoretically) optimal patch counts would be 1005, 1336, 1733, etc. Basically x^3 + 5.

Cheers, Joe

[shewhorn]

Just for my enlightenment, why would 4000 patches give a "terrible profile" over say 1100? (on a modern Epson or Canon LF)

The latest crop of printers are very linear and predictable in terms of their behavior. Because of this you can get away with using less patches and letting the software interpolate the values in between. In my experience this produces smoother gradients. Once you start increasing the patch count it starts to muck about with the linearity of the output and you can end up with areas of posterization in certain colors. The place where you might want to increase the patch count would be with older printers that are a little more unpredictable in terms of their output (or current printers that are unpredictable in their output) in attempt to accurate map and correct the inconsistencies in the device. Scott Martin or Andrew Rodney could probably explain it a bit better (and correct any errors if I've made any).

I wouldn't also mind some advice on dual filtered readings with iSis - how this works.

Kieth Cooper has a good writeup on this:

http://www.northlight-images.co.uk/reviews/profiling/isis_obc.html

Profilemaker is used in the article but the process carries over to i1Profiler.

I understand about OBA's and UV, filtered/unfiltered etc, but the practicalities of building a OBA-inclusive profile in i1Profiler on the double measurements has been beyond me so far.  For artworks shown in galleries with tungsten lighting, UV is minimal, but many galleries have some daylight, partially activating those OBAs, if present, so where do you draw the line?

Daylight or not, if the print was framed with archival considerations in mind the glass used will most likely block 99% of all UV anyway so... shine a black light on it, it won't make much of a difference. That said, not all artwork is shown in galleries! :-) Whether to use a paper with OBCs or not.... that seems to be a RAW vs. JPEG, Canon vs. Nikon, Mac vs. PC kind of thing. I'm not going to touch it! 

Cheers, Joe

[narikin]

Yes, that's correct. I was just pointing out that the patch generation is similar in multiples of 3 so if you use those patch sets you'll get similar results (and won't get the advantage of the additional patches that help with producing a more neutral grey). A few of the (theoretically) optimal patch counts would be 1005, 1336, 1733, etc. Basically x^3 + 5.

funny - thats the answer this thread was originally started for - a recommended patch number to start with!

The latest crop of printers are very linear and predictable in terms of their behavior. Because of this you can get away with using less patches and letting the software interpolate the values in between. In my experience this produces smoother gradients. Once you start increasing the patch count it starts to muck about with the linearity of the output and you can end up with areas of posterization in certain colors.

But surely an erroneous reading or two (spot of dust, micro ink drop out) would have greater effect, not corrected by averaging and throwing the profile 'off', with just 768 or 1005 patches vs 4000?

[stalisman]

If x is of the form 3n+1 then x^3 + 5 is divisible by 3  :-)

[shewhorn]

funny - thats the answer this thread was originally started for - a recommended patch number to start with!

I think people are hesitant to give a specific number because there are so many variables. Printer, ink, paper, and even the type of imagery you'll be printing will have an influence on what the optimal number of patches is. It would be a mistake to take those numbers and treat them with any kind of special regard. In theory it'll help with neutral grey tones. In reality it might make a difference, it might an imperceptible difference, and I'd put money down that there are cases out there where it could make things worse.

But surely an erroneous reading or two (spot of dust, micro ink drop out) would have greater effect, not corrected by averaging and throwing the profile 'off', with just 768 or 1005 patches vs 4000?

When I started learning more about building profiles I quickly discovered that my preconceived notions of what made perfectly logical sense were more often than not, wrong. The best approach (and I should have known this from my software engineering days) is to not go into it with specific expectations, or at the very least if you do, be completely open to the possibility that what you expect and what you see are going to be very different.

The BEST way to find out is to try it. Try a chart with 1005 patches and then bump it to 1006 patches. Try 1200, 1500, 1700, 4000, or even 6000 patch charts. Aside from costing you a bit in paper, you have an iSis which makes it really easy to read in charts of all sizes. If cost is a concern, get a cheap roll of paper to experiment with (but keep in mind that a different paper may need a different treatment).

Cheers, Joe